ADVERSARIAL PREDICTION OF RADIOTHERAPY TREATMENT PLANS

Abstract

Systems and methods are disclosed for generating radiotherapy treatment machine parameters based on projection images of a target anatomy. The systems and methods include operations including receiving a set of pairs of image data for each gantry angle of a radiotherapy treatment machine, wherein each pair of the set of pairs comprises a given projection image that represents a view of an anatomy of a subject from a given gantry angle and a given graphical aperture image of multi-leaf collimator (MLC) leaf positions at the given gantry angle based on the given projection image; training a generative adversarial network (GAN) model based on the set of pairs of image data for each gantry angle; and using the trained GAN model to predict an aperture image of MLC leaf positions for a desired gantry angle based on a projection image that represents a view of an anatomical region of interest.

Description

CLAIM FOR PRIORITY

This application claims the benefit of priority of U.S. application Ser. No. 16/784,919, filed Feb. 7, 2020, which is hereby incorporated by reference in its entirety.

TECHNICAL FIELD

Embodiments of the present disclosure pertain generally to determining machine parameters that direct the radiation therapy performed by a radiation therapy treatment system. In particular, the present disclosure pertains to using deep learning technologies to determine machine parameters that define a treatment plan in a radiation therapy system.

BACKGROUND

Radiation therapy (or “radiotherapy”) can be used to treat cancers or other ailments in mammalian (e.g., human and animal) tissue. One such radiotherapy technique is a Gamma Knife, by which a patient is irradiated by a large number of low-intensity gamma rays that converge with high intensity and high precision at a target (e.g., a tumor). In another embodiment, radiotherapy is provided using a linear accelerator, whereby a tumor is irradiated by high-energy particles (e.g., electrons, protons, ions, high-energy photons, and the like). The placement and dose of the radiation beam must be accurately controlled to ensure the tumor receives the prescribed radiation, and the placement of the beam should be such as to minimize damage to the surrounding healthy tissue, often called the organ(s) at risk (OARs). Radiation is termed “prescribed” because a physician orders a predefined amount of radiation to the tumor and surrounding organs similar to a prescription for medicine. Generally, ionizing radiation in the form of a collimated beam is directed from an external radiation source toward a patient.

A specified or selectable beam energy can be used, such as for delivering a diagnostic energy level range or a therapeutic energy level range. Modulation of a radiation beam can be provided by one or more attenuators or collimators (e.g., a multi-leaf collimator (MLC)). The intensity and shape of the radiation beam can be adjusted by collimation to avoid damaging healthy tissue (e.g., OARs) adjacent to the targeted tissue by conforming the projected beam to a profile of the targeted tissue.

The treatment planning procedure may include using a three-dimensional (3D) image of the patient to identify a target region (e.g., the tumor) and to identify critical organs near the tumor. Creation of a treatment plan can be a time-consuming process where a planner tries to comply with various treatment objectives or constraints (e.g., dose volume histogram (DVH), overlap volume histogram (OVH)), taking into account their individual importance (e.g., weighting) in order to produce a treatment plan that is clinically acceptable. This task can be a time-consuming trial-and-error process that is complicated by the various OARs because as the number of OARs increases (e.g., a dozen or more for a head-and-neck treatment), so does the complexity of the process. OARs distant from a tumor may be easily spared from radiation, while OARs close to or overlapping a target tumor may be difficult to spare.

Traditionally, for each patient, the initial treatment plan can be generated in an “offline” manner. The treatment plan can be developed well before radiation therapy is delivered, such as using one or more medical imaging techniques. Imaging information can include, for example, images from X-rays, computed tomography (CT), nuclear magnetic resonance (MR), positron emission tomography (PET), single-photon emission computed tomography (SPECT), or ultrasound. A health care provider, such as a physician, may use 3D imaging information indicative of the patient anatomy to identify one or more target tumors along with the OARs near the tumor(s). The health care provider can delineate the target tumor that is to receive a prescribed radiation dose using a manual technique, and the health care provider can similarly delineate nearby tissue, such as organs, at risk of damage from the radiation treatment. Alternatively or additionally, an automated tool (e.g., ABAS provided by Elekta AB, Sweden) can be used to assist in identifying or delineating the target tumor and organs at risk. A radiation therapy treatment plan (“treatment plan”) can then be created using numerical optimization techniques the minimize objective functions composed of clinical and dosimetric objectives and constraints (e.g., the maximum, minimum, and fraction of dose of radiation to a fraction of the tumor volume (“95% of target shall receive no less than 100% of prescribed dose”), and like measures for the critical organs). The optimized plan is comprised of numerical parameters that specify the direction, cross-sectional shape, and intensity of each radiation beam.

The treatment plan can then be later executed by positioning the patient in the treatment machine and delivering the prescribed radiation therapy directed by the optimized plan parameters. The radiation therapy treatment plan can include dose “fractioning,” whereby a sequence of radiation treatments is provided over a predetermined period of time (e.g., 30-45 daily fractions), with each treatment including a specified fraction of a total prescribed dose. However, during treatment, the position of the patient and the position of the target tumor in relation to the treatment machine (e.g., linear accelerator—“linac”) is very important in order to ensure the target tumor and not healthy tissue is irradiated.

Overview

In some embodiments, methods, systems and computer-readable medium are provided for generating one or more radiotherapy treatment plans. The methods, systems and computer-readable medium perform operations comprising: receiving a set of pairs of image data for each gantry angle of a radiotherapy treatment machine, wherein each pair of the set of pairs comprises a given projection image that represents a view of an anatomy of a subject from a given gantry angle and a given graphical aperture image of multi-leaf collimator (MLC) leaf positions at the given gantry angle based on the given projection image; training a generative adversarial network (GAN) model based on the set of pairs of image data for each gantry angle; and using the trained GAN model to predict an aperture image of MLC leaf positions for a desired gantry angle based on a projection image that represents a view of an anatomical region of interest.

In some implementations, the GAN comprises a conditional adversarial network (cGAN).

In some implementations, the GAN comprises a cycle-consistent generative adversarial network (CycleGAN).

In some implementations, the given gantry angle of a first pair of the set of pairs differs from the given gantry angle of a second pair of the set of pairs by a predetermined amount.

In some implementations, the GAN is configured to train the generative model using a discriminative model; values applied by the generative model and the discriminative model are established using adversarial training between the discriminative model and the generative model; and the generative model and the discriminative model comprise respective convolutional neural networks.

In some implementations, the adversarial training comprises: training the generative model to generate a first synthetic graphical aperture image representation of the MLC leaf positions at a first gantry angle from a projection image that represents a view of a training subject anatomy from the first gantry angle; and training the discriminative model to classify the first synthetic graphical aperture image as a synthetic or a real training example graphical aperture image; and an output of the generative model is used for training the discriminative model and an output of the discriminative model is used for training the generative model.

In some implementations, the GAN is trained using a cycle-consistent generative adversarial network (CycleGAN) comprising the generative model and the discriminative model, the generative model is a first generative model and the discriminative model is a first discriminative model, and the CycleGAN further comprises: a second generative model trained to: process, from a given pair of the set of pairs, a given graphical aperture image representation of the MLC leaf positions at a given gantry angle as an input; and provide a synthetic projection image that represents a view of a training subject anatomy from the given gantry angle as an output; and a second discriminative model trained to classify the synthetic projection image as a synthetic or a real projection image.

In some implementations, the CycleGAN comprises a first portion to train the first generative model, the first portion being trained to: obtain a set of training projection images representing different views of a patient anatomy from prior treatments that are paired with training graphical aperture images corresponding to each of the different views, each of the training graphical aperture images being aligned with a respective one of the training projection images; transmit the set of training projection images to an input of the first generative model to output a first set of graphical aperture images; receive the first set of graphical aperture images at an input of the first discriminative model to classify the first set of graphical aperture images as a synthetic or real training set of graphical aperture images; and receive the first set of graphical aperture images at an input of the second generative model to generate a first set of cycle projection images, for calculating cycle-consistency losses.

In some implementations, the CycleGAN comprises a second portion that is trained to: transmit the training graphical aperture images corresponding to each of the different views to the input of the second generative model to output a first set of synthetic projection images; receive the first set of synthetic projection images at an input of the second discriminative model to classify the first set of synthetic projection images as synthetic or real training projection images; and receive the first set of synthetic projection images at the input of the first generative model to generate a first set of cycle graphical aperture images for calculating cycle-consistency losses.

In some implementations, the cycle-consistency losses are generated based on a comparison of the first set of cycle projection images with the set of training projection images and a comparison of the first set of cycle graphical aperture images with the training graphical aperture images; the first generative model is trained to minimize or reduce a first loss term that represents an expectation of difference between a plurality of synthetic graphical aperture images and respectively paired training graphical aperture images; and the second generative model is trained to minimize or reduce a second loss term that represents an expectation of difference between a plurality of synthetic projection images and respectively paired training projection images.

In some implementations, the operations comprise: obtaining radiotherapy treatment machine parameter information representing MLC leaf positions at gantry angles corresponding to each of the views represented by the projection images of the set of pairs of image data and radiotherapy beam intensities corresponding to each of the projection images; generating training graphical aperture image representations based on the obtained radiotherapy treatment machine parameter information; and aligning each of the generated training graphical aperture image representations with the corresponding projection images.

In some implementations, the training graphical aperture image representations and corresponding projection images are two-dimensional images or three-dimensional images comprising stacks of two-dimensional projection image and graphical aperture image pairs corresponding to an entire treatment fraction.

In some implementations, the operations comprise generating one or more radiotherapy treatment machine parameters from the aperture image predicted by the trained GAN model.

In some implementations, the operations comprise computing a dose volume histogram or a three-dimensional dose distribution based on the generated one or more radiotherapy treatment machine parameters.

In some implementations, the one or more parameters include at least one of a gantry angle, an MLC jaw position, an MLC leaf position, or a radiation therapy beam intensity.

In some implementations, the operations comprise computing dosimetric parameters based on the generated one or more radiotherapy treatment machine parameters.

In some implementations, the given projection images in the set of pairs of image data is generated by ray tracing or Fourier reconstruction.

In some implementations, the operations comprise generating, as the projection images in the set of pairs of image data, a plurality of two-dimensional projection images that represent a plurality of two-dimensional views of the anatomy from a plurality of different gantry angles.

In some implementations, the plurality of two-dimensional projection images represents views of the anatomy at a plurality of angles spanning the arc of motion of a gantry of the radiotherapy treatment machine during therapy.

The above overview is intended to provide an overview of subject matter of the present patent application. It is not intended to provide an exclusive or exhaustive explanation of the inventive subject matter. The detailed description is included to provide further information about the present patent application.

BRIEF DESCRIPTION OF THE DRAWINGS

In the drawings, which are not necessarily drawn to scale, like numerals describe substantially similar components throughout the several views. Like numerals having different letter suffixes represent different instances of substantially similar components. The drawings illustrate generally, by way of example but not by way of limitation, various embodiments discussed in the present document.

FIG. 1 illustrates an exemplary radiotherapy system, according to some embodiments of the present disclosure.

FIG. 2A illustrates an exemplary radiation therapy system that can include radiation therapy output configured to provide a therapy beam, according to some embodiments of the present disclosure.

FIG. 2B illustrates an exemplary system including a combined radiation therapy system and an imaging system, such as a cone beam computed tomography (CBCT) imaging system, according to some embodiments of the present disclosure.

FIG. 3 illustrates a partially cut-away view of an exemplary system including a combined radiation therapy system and an imaging system, such as a nuclear magnetic resonance (MR) imaging (MRI) system, according to some embodiments of the present disclosure.

FIGS. 4A and 4B depict the differences between an exemplary MRI image and a corresponding CT image, respectively, according to some embodiments of the present disclosure.

FIG. 5 illustrates an exemplary collimator configuration for shaping, directing, or modulating an intensity of a radiation therapy beam, according to some embodiments of the present disclosure.

FIG. 6 illustrates an exemplary Gamma Knife radiation therapy system, according to some embodiments of the present disclosure.

FIGS. 7A-D illustrate exemplary flow diagrams for deep learning, according to some embodiments of the present disclosure.

FIGS. 8A and 8B illustrate projection views from various directions, according to some embodiments of the present disclosure.

FIG. 9 illustrates an exemplary method for generating training data for deep learning, according to some embodiments of the present disclosure.

FIG. 10 illustrates an exemplary method for training deep learning using projection images and graphical representations of leaf positions, according to some embodiments of the present disclosure.

FIGS. 11A and 11B illustrate exemplary data flows for training and use of a machine learning model to generate radiotherapy equipment parameters according to some embodiments of the present disclosure.

FIG. 12 illustrates a method for using trained deep learning to generate radiotherapy equipment parameters from a projection image of a given view of anatomy from a given gantry angle, according to some embodiments of the present disclosure.

FIG. 13 illustrates an exemplary block diagram of a machine on which one or more of the methods as discussed herein can be implemented.

DETAILED DESCRIPTION

In the following detailed description, reference is made to the accompanying drawings which form a part hereof, and which is shown by way of illustration-specific embodiments in which the present disclosure may be practiced. These embodiments, which are also referred to herein as “examples,” are described in sufficient detail to enable those skilled in the art to practice the disclosure, and it is to be understood that the embodiments may be combined, or that other embodiments may be utilized and that structural, logical and electrical changes may be made without departing from the scope of the present disclosure. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present disclosure is defined by the appended claims and their equivalents.

Intensity modulated radiotherapy (IMRT) and volumetric modulated arc therapy (VMAT) have become the standards of care in modern cancer radiation therapy. Creating individual patient IMRT or VMAT treatment plans is a trial-and-error process, weighing target dose versus OAR sparing tradeoffs, and adjusting program constraints whose effects on the plan quality metrics and the dose distribution can be very difficult to anticipate. Indeed, the order in which the planning constraints are adjusted can itself result in dose differences. Treatment plan quality depends on often subjective judgements by the planner that depend on his/her experience and skill. Even the most skilled planners still have no assurance that their plans are close to the best possible, or whether a little or a lot of effort will result in a significantly better plan.

The present disclosure includes various techniques to improve and enhance radiotherapy treatment by predicting radiotherapy treatment machine parameters (also referred to as control points) for delivering intended dose distributions based on a patient's delineated anatomy. The technical benefits include reduced radiotherapy treatment plan creation time and may result in many apparent medical treatment benefits (including improved accuracy of radiotherapy treatment, reduced exposure to unintended radiation, and the like). The disclosed techniques may be applicable to a variety of medical treatment and diagnostic settings or radiotherapy treatment equipment and devices.

The present disclosure predicts or estimates control points by training a machine learning model, such as a generative adversarial network (GAN), on training data consisting of pairs of control points (represented by graphical aperture images) and the corresponding images of a patient (represented by projection images). These data pairs are constructed and aligned so that each element of the image data corresponds to an element of the control point data. The disclosed machine learning method then determines a mapping or relationship between the two data realms. This learned model associates the image data and the control point data without explicit reference to the physics of radiation transport or to the many details attending the creation of treatment plans from which the training data were extracted.

According to some embodiments, the disclosure directly learns (trains a machine learning model based on) the parameters of a population of exemplary treatment plans for a common diagnosis, and then uses that model to predict the treatment machine parameter settings required to deliver the intended dose distributions based only on the new patient's delineated anatomy. The treatment parameters include the linac gantry angles, beam aperture shapes through which the therapeutic radiation beams are projected at the target and the beams' intensities, any parameter used to delivery electron or particle therapy, among others. This disclosure estimates the machine parameters to drive the treatment machine, given the image of the patient and the target and OAR delineations.

According to some embodiments, a kind of statistical learning based on deep neural networks (deep learning) is used to obtain a much more detailed model of the linkages between patient anatomies and treatment constraints and the resulting treatment machine parameter settings that would deliver the intended 3D dose distribution. Such a result is impossible with any of the earlier plan quality analyses. By employing a machine learning model, this disclosure produces a model of the treatment planning process, encapsulating the many (subjective) decisions made during plan creation and enables: the production of plan templates to initiate plan creation, the assessment of existing plans' likely quality, aid for treatment clinics lacking deep local IMRT/VMAT expertise, and fully automated treatment planning. By using deep learning to generate control points of a radiotherapy device and the implied treatment plans, computational complexity of the total plan optimization process is reduced and the time needed to create a treatment plan for a given patient is reduced.

Specifically, according to some embodiments, a machine learning model is trained to establish a relationship between a projection of a given image of a patient anatomy corresponding to a given angle of a gantry of the radiotherapy device and a graphical aperture image representation of MLC leaf positions of the radiotherapy device. The machine learning model is trained by obtaining the projection images representing different views of a patient anatomy from prior treatments; and for each of the prior treatments: obtaining control point information representing MLC leaf positions at gantry angles corresponding to each of the different views and radiotherapy beam intensity corresponding to each of the projection images to generate ground truth (or training) graphical aperture image representations based on the obtained control point information paired with the projection images. Subsequently, one or more parameters of the machine learning model are adjusted based on the paired generated training graphical aperture image representations and corresponding projection images. Once trained, a new patient projection image is obtained and a corresponding graphical aperture image representation of MLC leaf positions of the radiotherapy device are estimated using the trained machine learning model. Other embodiments may include graphical representations of the beams' intensities (e.g., bar graphs) in addition to the MLC leaf positions.

In some embodiments, control points of the radiotherapy device are then computed based on the estimated graphical aperture image representation of MLC leaf positions. Namely, a reverse mapping function is utilized to generate control points (e.g., beam intensity, gantry angle, and/or MLC leaf positions of individual leaves of the MLC) to achieve a beam that corresponds to the estimated graphical aperture image representation of the MLC leaf positions. In particular, the graphical aperture image representation of MLC leaf positions identifies a resulting beam shape that is output by the radiotherapy device and the reverse mapping function is utilized to generate the control points to provide the resulting beam shape that is represented by the graphical aperture image representation of MLC leaf positions.

FIG. 1 illustrates an exemplary radiotherapy system 100 for providing radiation therapy to a patient. The radiotherapy system 100 includes an image processing device 112. The image processing device 112 may be connected to a network 120. The network 120 may be connected to the Internet 122. The network 120 can connect the image processing device 112 with one or more of a database 124, a hospital database 126, an oncology information system (OIS) 128, a radiation therapy device 130, an image acquisition device 132, a display device 134, and a user interface 136. The image processing device 112 can be configured to generate radiation therapy treatment plans 142 to be used by the radiation therapy device 130.

The image processing device 112 may include a memory device 116, an image processor 114, and a communication interface 118. The memory device 116 may store computer-executable instructions, such as an operating system 143, radiation therapy treatment plans 142 (e.g., original treatment plans, adapted treatment plans and the like), software programs 144 (e.g., artificial intelligence, deep learning, neural networks, radiotherapy treatment plan software), and any other computer-executable instructions to be executed by the processor 114. In one embodiment, the software programs 144 may convert medical images of one format (e.g., MRI) to another format (e.g., CT) by producing synthetic images, such as pseudo-CT images. For instance, the software programs 144 may include image processing programs to train a predictive model for converting a medical image 146 in one modality (e.g., an MRI image) into a synthetic image of a different modality (e.g., a pseudo CT image); alternatively, the trained predictive model may convert a CT image into an MRI image. In another embodiment, the software programs 144 may register the patient image (e.g., a CT image or an MR image) with that patient's dose distribution (also represented as an image) so that corresponding image voxels and dose voxels are associated appropriately by the network. In yet another embodiment, the software programs 144 may substitute functions of the patient images such as signed distance functions or processed versions of the images that emphasize some aspect of the image information. Such functions might emphasize edges or differences in voxel textures, or any other structural aspect useful to neural network learning. In another embodiment, the software programs 144 may substitute functions of the dose distribution that emphasize some aspect of the dose information. Such functions might emphasize steep gradients around the target or any other structural aspect useful to neural network learning. The memory device 116 may store data, including medical images 146, patient data 145, and other data required to create and implement a radiation therapy treatment plan 142.

In yet another embodiment, the software programs 144 may generate projection images for a set of two-dimensional (2D) and/or 3D CT or MR images depicting an anatomy (e.g., one or more targets and one or more OARs) representing different views of the anatomy from one or more gantry angles of the radiotherapy equipment. For example, the software programs 144 may process the set of CT or MR images and create a stack of projection images depicting different views of the anatomy depicted in the CT or MR images from various perspectives of the gantry of the radiotherapy equipment. In particular, one projection image may represent a view of the anatomy from 0 degrees of the gantry, a second projection image may represent a view of the anatomy from 45 degrees of the gantry, and a third projection image may represent a view of the anatomy from 90 degrees of the gantry. The degrees may be a position of the MLC relative to a particular axis of the anatomy depicted in the CT or MR images. The axis may remain the same for each of the different degrees that are measured.

Projection views for a simple ellipse 802 are shown schematically in FIG. 8A. In FIG. 8A, the views are oriented relative the ellipse center and capture the shape and extent of the ellipse 802 as seen from each angle (e.g., 0 degrees represented by view 803, 45 degrees represented by view 804, and 90 degrees represented by view 805). For example, the view of ellipse 802 when seen from a 0-degree angle relative to the y-axis 807 of ellipse 802 is projected as view 803. For example, the view of ellipse 802 when seen from a 45-degree angle relative to the y-axis 807 of ellipse 802 is projected as view 804. For example, the view of ellipse 802 when seen from a 90-degree angle relative to the y-axis 807 of ellipse 802 is projected as view 805.

Projections of the male pelvic anatomy relative to a set of original 3D CT images 801 are shown in FIG. 8B. Selected organs at risk and target organs were contoured in the 3D CT image 801 and their voxels were assigned a constant value. Projection images 850 at selected angles (0 degrees, 45 degrees, and 90 degrees) about the central axis of the 3D CT image 801 can be obtained using the forward projection capability of a cone beam CT reconstruction program. Projection images can also be computed either by directly re-creating the projection view geometry by ray tracing or by Fourier reconstruction as in computed tomography.

For example, the projection image can be computed by tracing the path of light as pixels in an image plane and simulating the effects of its encounters with virtual objects. In some implementations, the projection image is generated by tracing a path from an imaginary eye (the MLC view or beam's eye view) through each pixel in a virtual screen and calculating the color of the object visible through it. Any other tomographic reconstruction technique can be utilized to generate the projection images from the MLC beam's eye view of the anatomy depicted in the 3D CT images 801.

For example, the set of (or collection of) 3D CT images 801 can be used to generate one or more views of the anatomy (e.g., the bladder, prostate, seminal vesicles, rectum, first and second targets) depicted in the 3D CT images 801. The views can be from the perspective of the MLC including the gantry of the radiotherapy device and, for simplicity with reference to FIG. 8B, the views are measured in degrees relative to the y-axis of the 3D CT images 801 and based on a distance between the anatomy depicted in the image and the MLC. Specifically, a first view 810 represents a projection of the 3D CT images 801 when viewed or seen by the gantry when the gantry is 0 degrees relative to the y-axis and is at a given distance from the anatomy depicted in the 3D CT image 801, a second view 820 represents a projection of the 3D CT images 801 when viewed or seen by the gantry when the gantry is 45 degrees relative to the y-axis and is at a given distance from the anatomy depicted in the 3D CT image 801, and a third view 830 represents a projection of the 3D CT images 801 when viewed or seen by the gantry when the gantry is 90 degrees relative to the y-axis. Any other views can be provided, such as a different view at each of 360 degrees around the anatomy depicted in the 3D CT images 801.

Referring back to FIG. 1, in yet another embodiment, the software programs 144 may generate graphical aperture image representations of MLC leaf positions (sometimes referred to as graphical aperture images) at various gantry angles. These graphical aperture images are also referred to as aperture images or training graphical aperture images. In particular, the software programs 144 may receive a set of control points that are used to control a radiotherapy device to produce a radiotherapy beam. The control points may represent the beam intensity, gantry angle relative to the patient position, and the leaf positions of the MLC, among other machine parameters. Based on these control points, a graphical image may be generated to graphically represent the beam shape and intensity that is output by the MLC at each particular gantry angle. The software programs 144 may align and pair each graphical image of the aperture at a particular gantry angle with the corresponding projection image at that angle that was generated. The images are aligned and scaled with the projections such that each projection image pixel is aligned with the corresponding aperture image pixel.

FIG. 9 illustrates an exemplary method for generating training data for deep learning, according to some embodiments of the present disclosure. Specifically, in FIG. 9 images and control point parameters are transformed into 3D image volumes. In the top portion of FIG. 9, 3D CT images 801 are converted to a stack of projection images 850 at selected angles about the central axis of the 3D CT image 801 in a similar manner as discussed in connection with FIGS. 8A and 8B. In the lower portion of FIG. 9, the control point angles 910, apertures, and intensities, corresponding to each angle of the gantry represented by the projection images 850, are recreated as graphical aperture images 920. The shaded portion 922 shows the openings between MLC left and right leaf edges that permit radiation to pass in each graphical aperture image 920. Graphical aperture images 920 are aligned and scaled with the projection images 850 such that each projection pixel is aligned with the corresponding aperture pixel that irradiates it. In some embodiment, a bar object can be presented for each graphical aperture image (e.g., at the lower left of the graphical aperture image 920). The length of this bar object encodes the radiation intensity at this control point and is estimated along with the leaf positions.

The control point parameters 910 (Y={ϕ, ( . . . L_n^ϕ, R_n^φ . . . ), y^φ}_k, k=1, . . . , K) represent the gantry angles, the MLC apertures at each gantry angle, and the radiation intensity at that angle. The apertures are depicted as graphical aperture images 920. Each graphical aperture image 920 is assigned and paired with a corresponding one of the anatomy projection images 850 at the same gantry angle. Each shaded portion 922 represents an opening between pairs of opposing tungsten MLC leaves (e.g., 532A-J) and it is these apertures that control the shape of the X-ray beam to cover the target to the prescribed radiation dose. The projection images 850 and the graphical aperture images 920 are scaled and aligned to ensure that each anatomy pixel in the projection images 850 is aligned with the corresponding aperture pixel irradiating that anatomy element. The construction of anatomy and control point data are represented as aligned 3D image volumes with common dimensions, pixel spacing, and origin.

Referring back to FIG. 1, in yet another embodiment, the software programs 144 store a treatment planning software that includes a trained machine learning model, such as a trained GAN or trained CycleGAN, to generate or estimate a synthetic graphical aperture image representation of MLC leaf positions at a given gantry angle for a projection image of the anatomy representing the view of the anatomy from the given gantry angle. The software programs 144 may further store a function to convert or compute machine parameters or control points for a given type of machine to output a beam from the MLC that achieves the same or similar estimated graphical aperture image representation of the MLC leaf positions. Namely, the treatment planning software may output an image representing an estimated image of the beam shape and intensity for a given gantry angle and for a given projection image of the gantry at that angle, and the function may compute the control points for a given radiotherapy device to achieve that beam shape and intensity.

In addition to the memory device 116 storing the software programs 144, it is contemplated that software programs 144 may be stored on a removable computer medium, such as a hard drive, a computer disk, a CD-ROM, a DVD, a HD, a Blu-Ray DVD, USB flash drive, a SD card, a memory stick, or any other suitable medium; and the software programs 144 when downloaded to image processing device 112 may be executed by image processor 114.

The processor 114 may be communicatively coupled to the memory device 116, and the processor 114 may be configured to execute computer-executable instructions stored thereon. The processor 114 may send or receive medical images 146 to memory device 116. For example, the processor 114 may receive medical images 146 from the image acquisition device 132 via the communication interface 118 and network 120 to be stored in memory device 116. The processor 114 may also send medical images 146 stored in memory device 116 via the communication interface 118 to the network 120 to be either stored in database 124 or the hospital database 126.

Further, the processor 114 may utilize software programs 144 (e.g., a treatment planning software) along with the medical images 146 and patient data 145 to create the radiation therapy treatment plan 142. Medical images 146 may include information such as imaging data associated with a patient anatomical region, organ, or volume of interest segmentation data. Patient data 145 may include information such as (1) functional organ modeling data (e.g., serial versus parallel organs, appropriate dose response models, etc.); (2) radiation dosage data (e.g., DVH information; or (3) other clinical information about the patient and course of treatment (e.g., other surgeries, chemotherapy, previous radiotherapy, etc.).

In addition, the processor 114 may utilize software programs to generate intermediate data such as updated parameters to be used, for example, by a machine learning model, such as a neural network model, or generate intermediate 2D or 3D images, which may then subsequently be stored in memory device 116. The processor 114 may subsequently then transmit the executable radiation therapy treatment plan 142 via the communication interface 118 to the network 120 to the radiation therapy device 130, where the radiation therapy plan will be used to treat a patient with radiation. In addition, the processor 114 may execute software programs 144 to implement functions such as image conversion, image segmentation, deep learning, neural networks, and artificial intelligence. For instance, the processor 114 may execute software programs 144 that train or contour a medical image: such software programs 144 when executed may train a boundary detector or utilize a shape dictionary.

The processor 114 may be a processing device, include one or more general-purpose processing devices such as a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), or the like. More particularly, the processor 114 may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction Word (VLIW) microprocessor, a processor implementing other instruction sets, or processors implementing a combination of instruction sets. The processor 114 may also be implemented by one or more special-purpose processing devices such as an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a System on a Chip (SoC), or the like. As would be appreciated by those skilled in the art, in some embodiments, the processor 114 may be a special-purpose processor, rather than a general-purpose processor. The processor 114 may include one or more known processing devices, such as a microprocessor from the Pentium™, Core™, Xeon™, or Itanium® family manufactured by Intel™, the Turion™, Athlon™, Sempron™, Opteron™, FX™, Phenom™ family manufactured by AMDT™, or any of various processors manufactured by Sun Microsystems. The processor 114 may also include graphical processing units such as a GPU from the GeForce®, Quadro®, Tesla® family manufactured by Nvidia™, GMA, Iris™ family manufactured by Intel™, or the Radeon™ family manufactured by AMD™. The processor 114 may also include accelerated processing units such as the Xeon Phi™ family manufactured by Intel™ The disclosed embodiments are not limited to any type of processor(s) otherwise configured to meet the computing demands of identifying, analyzing, maintaining, generating, and/or providing large amounts of data or manipulating such data to perform the methods disclosed herein. In addition, the term “processor” may include more than one processor (for example, a multi-core design or a plurality of processors each having a multi-core design). The processor 114 can execute sequences of computer program instructions, stored in memory device 116, to perform various operations, processes, methods that will be explained in greater detail below.

The memory device 116 can store medical images 146. In some embodiments, the medical images 146 may include one or more MRI images (e.g., 2D MRI, 3D MRI, 2D streaming MRI, four-dimensional (4D) MRI, 4D volumetric MRI, 4D cine MRI, projection images, graphical aperture images, and pairing information between projection images and graphical aperture images, etc.), functional MRI images (e.g., fMRI, DCE-MRI, diffusion MRI), CT images (e.g., 2D CT, cone beam CT, 3D CT, 4D CT), ultrasound images (e.g., 2D ultrasound, 3D ultrasound, 4D ultrasound), one or more projection images representing views of an anatomy depicted in the MRI, synthetic CT (pseudo-CT), and/or CT images at different angles of a gantry relative to a patient axis, PET images, X-ray images, fluoroscopic images, radiotherapy portal images, SPECT images, computer generated synthetic images (e.g., pseudo-CT images), aperture images, graphical aperture image representations of MLC leaf positions at different gantry angles, and the like. Further, the medical images 146 may also include medical image data, for instance, training images, and training images, contoured images, and dose images. In an embodiment, the medical images 146 may be received from the image acquisition device 132. Accordingly, image acquisition device 132 may include an MRI imaging device, a CT imaging device, a PET imaging device, an ultrasound imaging device, a fluoroscopic device, a SPECT imaging device, an integrated linac and MRI imaging device, or other medical imaging devices for obtaining the medical images of the patient. The medical images 146 may be received and stored in any type of data or any type of format that the image processing device 112 may use to perform operations consistent with the disclosed embodiments.

The memory device 116 may be a non-transitory computer-readable medium, such as a read-only memory (ROM), a phase-change random access memory (PRAM), a static random access memory (SRAM), a flash memory, a random access memory (RAM), a dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM), an electrically erasable programmable read-only memory (EEPROM), a static memory (e.g., flash memory, flash disk, static random access memory) as well as other types of random access memories, a cache, a register, a CD-ROM, a DVD or other optical storage, a cassette tape, other magnetic storage device, or any other non-transitory medium that may be used to store information including image, data, or computer executable instructions (e.g., stored in any format) capable of being accessed by the processor 114, or any other type of computer device. The computer program instructions can be accessed by the processor 114, read from the ROM, or any other suitable memory location, and loaded into the RAM for execution by the processor 114. For example, the memory device 116 may store one or more software applications. Software applications stored in the memory device 116 may include, for example, an operating system 143 for common computer systems as well as for software-controlled devices. Further, the memory device 116 may store an entire software application, or only a part of a software application, that are executable by the processor 114. For example, the memory device 116 may store one or more radiation therapy treatment plans 142.

The image processing device 112 can communicate with the network 120 via the communication interface 118, which can be communicatively coupled to the processor 114 and the memory device 116. The communication interface 118 may provide communication connections between the image processing device 112 and radiotherapy system 100 components (e.g., permitting the exchange of data with external devices). For instance, the communication interface 118 may in some embodiments have appropriate interfacing circuitry to connect to the user interface 136, which may be a hardware keyboard, a keypad, or a touch screen through which a user may input information into radiotherapy system 100.

Communication interface 118 may include, for example, a network adaptor, a cable connector, a serial connector, a USB connector, a parallel connector, a high-speed data transmission adaptor (e.g., such as fiber, USB 3.0, thunderbolt, and the like), a wireless network adaptor (e.g., such as a WiFi adaptor), a telecommunication adaptor (e.g., 3G, 4G/LTE and the like), and the like. Communication interface 118 may include one or more digital and/or analog communication devices that permit image processing device 112 to communicate with other machines and devices, such as remotely located components, via the network 120.

The network 120 may provide the functionality of a local area network (LAN), a wireless network, a cloud computing environment (e.g., software as a service, platform as a service, infrastructure as a service, etc.), a client-server, a wide area network (WAN), and the like. For example, network 120 may be a LAN or a WAN that may include other systems S1 (138), S2 (140), and S3 (141). Systems S1, S2, and S3 may be identical to image processing device 112 or may be different systems. In some embodiments, one or more of the systems in network 120 may form a distributed computing/simulation environment that collaboratively performs the embodiments described herein. In some embodiments, one or more systems S1, S2, and S3 may include a CT scanner that obtains CT images (e.g., medical images 146). In addition, network 120 may be connected to Internet 122 to communicate with servers and clients that reside remotely on the Internet.

Therefore, network 120 can allow data transmission between the image processing device 112 and a number of various other systems and devices, such as the OIS 128, the radiation therapy device 130, and the image acquisition device 132. Further, data generated by the OIS 128 and/or the image acquisition device 132 may be stored in the memory device 116, the database 124, and/or the hospital database 126. The data may be transmitted/received via network 120, through communication interface 118 in order to be accessed by the processor 114, as required.

The image processing device 112 may communicate with database 124 through network 120 to send/receive a plurality of various types of data stored on database 124. For example, database 124 may include machine data (control points) that includes information associated with a radiation therapy device 130, image acquisition device 132, or other machines relevant to radiotherapy. Machine data information may include control points, such as radiation beam size, arc placement, beam on and off time duration, machine parameters, segments, MLC configuration, gantry speed, MRI pulse sequence, and the like. Database 124 may be a storage device and may be equipped with appropriate database administration software programs. One skilled in the art would appreciate that database 124 may include a plurality of devices located either in a central or a distributed manner.

In some embodiments, database 124 may include a processor-readable storage medium (not shown). While the processor-readable storage medium in an embodiment may be a single medium, the term “processor-readable storage medium” should be taken to include a single medium or multiple media (e.g., a centralized or distributed database, and/or associated caches and servers) that store the one or more sets of computer-executable instructions or data. The term “processor-readable storage medium” shall also be taken to include any medium that is capable of storing or encoding a set of instructions for execution by a processor and that cause the processor to perform any one or more of the methodologies of the present disclosure. The term “processor-readable storage medium” shall accordingly be taken to include, but not be limited to, solid-state memories, optical and magnetic media. For example, the processor-readable storage medium can be one or more volatile, non-transitory, or non-volatile tangible computer-readable media.

Image processor 114 may communicate with database 124 to read images into memory device 116 or store images from memory device 116 to database 124. For example, the database 124 may be configured to store a plurality of images (e.g., 3D MRI, 4D MRI, 2D MRI slice images, CT images, 2D Fluoroscopy images, X-ray images, raw data from MR scans or CT scans, Digital Imaging and Communications in Medicine (DIMCOM) data, projection images, graphical aperture images, etc.) that the database 124 received from image acquisition device 132. Database 124 may store data to be used by the image processor 114 when executing software program 144, or when creating radiation therapy treatment plans 142. Database 124 may store the data produced by the trained machine leaning mode, such as a neural network including the network parameters constituting the model learned by the network and the resulting predicted data. The image processing device 112 may receive the imaging data, such as a medical image 146 (e.g., 2D MRI slice images, CT images, 2D Fluoroscopy images, X-ray images, 3DMRI images, 4D MRI images, projection images, graphical aperture images, etc.) either from the database 124, the radiation therapy device 130 (e.g., an MRI-linac), and/or the image acquisition device 132 to generate a treatment plan 142.

In an embodiment, the radiotherapy system 100 can include an image acquisition device 132 that can acquire medical images (e.g., MRI images, 3D MRI, 2D streaming MRI, 4D volumetric MRL, CT images, cone-Beam CT, PET images, functional MRI images (e.g., fMRI, DCE-MRI and diffusion MRI), X-ray images, fluoroscopic image, ultrasound images, radiotherapy portal images, SPECT images, and the like) of the patient. Image acquisition device 132 may, for example, be an MRI imaging device, a CT imaging device, a PET imaging device, an ultrasound device, a fluoroscopic device, a SPECT imaging device, or any other suitable medical imaging device for obtaining one or more medical images of the patient. Images acquired by the image acquisition device 132 can be stored within database 124 as either imaging data and/or test data. By way of example, the images acquired by the image acquisition device 132 can also be stored by the image processing device 112, as medical images 146 in memory device 116.

In an embodiment, for example, the image acquisition device 132 may be integrated with the radiation therapy device 130 as a single apparatus (e.g., an MRI-linac). Such an MRI-linac can be used, for example, to determine a location of a target organ or a target tumor in the patient, so as to direct radiation therapy accurately according to the radiation therapy treatment plan 142 to a predetermined target.

The image acquisition device 132 can be configured to acquire one or more images of the patient's anatomy for a region of interest (e.g., a target organ, a target tumor, or both). Each image, typically a 2D image or slice, can include one or more parameters (e.g., a 2D slice thickness, an orientation, and a location, etc.). In an embodiment, the image acquisition device 132 can acquire a 2D slice in any orientation. For example, an orientation of the 2D slice can include a sagittal orientation, a coronal orientation, or an axial orientation. The processor 114 can adjust one or more parameters, such as the thickness and/or orientation of the 2D slice, to include the target organ and/or target tumor. In an embodiment, 2D slices can be determined from information such as a 3D MRI volume. Such 2D slices can be acquired by the image acquisition device 132 in “real-time” while a patient is undergoing radiation therapy treatment, for example, when using the radiation therapy device 130, with “real-time” meaning acquiring the data in at least milliseconds or less.

The image processing device 112 may generate and store radiation therapy treatment plans 142 for one or more patients. The radiation therapy treatment plans 142 may provide information about a particular radiation dose to be applied to each patient. The radiation therapy treatment plans 142 may also include other radiotherapy information, such as control points including beam angles, gantry angles, beam intensity, dose-histogram-volume information, the number of radiation beams to be used during therapy, the dose per beam, and the like.

The image processor 114 may generate the radiation therapy treatment plan 142 by using software programs 144 such as treatment planning software (such as Monaco®, manufactured by Elekta AB of Stockholm, Sweden). In order to generate the radiation therapy treatment plans 142, the image processor 114 may communicate with the image acquisition device 132 (e.g., a CT device, an MRI device, a PET device, an X-ray device, an ultrasound device, etc.) to access images of the patient and to delineate a target, such as a tumor. In some embodiments, the delineation of one or more OARs, such as healthy tissue surrounding the tumor or in close proximity to the tumor may be required. Therefore, segmentation of the OAR may be performed when the OAR is close to the target tumor. In addition, if the target tumor is close to the OAR (e.g., prostate in near proximity to the bladder and rectum), then by segmenting the OAR from the tumor, the radiotherapy system 100 may study the dose distribution not only in the target but also in the OAR.

In order to delineate a target organ or a target tumor from the OAR, medical images, such as MIRI images, CT images, PET images, fMRI images, X-ray images, ultrasound images, radiotherapy portal images, SPECT images, and the like, of the patient undergoing radiotherapy may be obtained non-invasively by the image acquisition device 132 to reveal the internal structure of a body part. Based on the information from the medical images, a 3D structure of the relevant anatomical portion may be obtained. In addition, during a treatment planning process, many parameters may be taken into consideration to achieve a balance between efficient treatment of the target tumor (e.g., such that the target tumor receives enough radiation dose for an effective therapy) and low irradiation of the OAR(s) (e.g., the OAR(s) receives as low a radiation dose as possible). Other parameters that may be considered include the location of the target organ and the target tumor, the location of the OAR, and the movement of the target in relation to the OAR. For example, the 3D structure may be obtained by contouring the target or contouring the OAR within each 2D layer or slice of an MRI or CT image and combining the contour of each 2D layer or slice. The contour may be generated manually (e.g., by a physician, dosimetrist, or health care worker using a program such as MONACO™ manufactured by Elekta AB of Stockholm, Sweden) or automatically (e.g., using a program such as the Atlas-based auto-segmentation software, ABAS™, and a successor auto-segmentation software product ADMIRE™, manufactured by Elekta AB of Stockholm, Sweden). In certain embodiments, the 3D structure of a target tumor or an OAR may be generated automatically by the treatment planning software.

After the target tumor and the OAR(s) have been located and delineated, a dosimetrist, physician, or healthcare worker may determine a dose of radiation to be applied to the target tumor, as well as any maximum amounts of dose that may be received by the OAR proximate to the tumor (e.g., left and right parotid, optic nerves, eyes, lens, inner ears, spinal cord, brain stem, and the like). After the radiation dose is determined for each anatomical structure (e.g., target tumor, OAR), a process known as inverse planning may be performed to determine one or more treatment plan parameters that would achieve the desired radiation dose distribution. Examples of treatment plan parameters include volume delineation parameters (e.g., which define target volumes, contour sensitive structures, etc.), margins around the target tumor and OARs, beam angle selection, collimator settings, and beam-on times. During the inverse-planning process, the physician may define dose constraint parameters that set bounds on how much radiation an OAR may receive (e.g., defining full dose to the tumor target and zero dose to any OAR; defining 95% of dose to the target tumor; defining that the spinal cord, brain stem, and optic structures receive ≤45Gy, ≤55Gy and <54Gy, respectively). The result of inverse planning may constitute a radiation therapy treatment plan 142 that may be stored in memory device 116 or database 124. Some of these treatment parameters may be correlated. For example, tuning one parameter (e.g., weights for different objectives, such as increasing the dose to the target tumor) in an attempt to change the treatment plan may affect at least one other parameter, which in turn may result in the development of a different treatment plan. Thus, the image processing device 112 can generate a tailored radiation therapy treatment plan 142 having these parameters in order for the radiation therapy device 130 to provide radiotherapy treatment to the patient.

In addition, the radiotherapy system 100 may include a display device 134 and a user interface 136. The display device 134 may include one or more display screens that display medical images, interface information, treatment planning parameters (e.g., projection images, graphical aperture images, contours, dosages, beam angles, etc.) treatment plans, a target, localizing a target and/or tracking a target, or any related information to the user. The user interface 136 may be a keyboard, a keypad, a touch screen or any type of device with which a user may input information to radiotherapy system 100. Alternatively, the display device 134 and the user interface 136 may be integrated into a device such as a tablet computer (e.g., Apple iPad®, Lenovo Thinkpad®, Samsung Galaxy®, etc.).

Furthermore, any and all components of the radiotherapy system 100 may be implemented as a virtual machine (e.g., VMWare, Hyper-V, and the like). For instance, a virtual machine can be software that functions as hardware. Therefore, a virtual machine can include at least one or more virtual processors, one or more virtual memories, and one or more virtual communication interfaces that together function as hardware. For example, the image processing device 112, the OIS 128, the image acquisition device 132 could be implemented as a virtual machine Given the processing power, memory, and computational capability available, the entire radiotherapy system 100 could be implemented as a virtual machine.

FIG. 2A illustrates an exemplary radiation therapy device 202 that may include a radiation source, such as an X-ray source or a linear accelerator, a couch 216, an imaging detector 214, and a radiation therapy output 204. The radiation therapy device 202 may be configured to emit a radiation beam 208 to provide therapy to a patient. The radiation therapy output 204 can include one or more attenuators or collimators, such as an MLC as described in the illustrative embodiment of FIG. 5, below.

Referring back to FIG. 2A, a patient can be positioned in a region 212 and supported by the treatment couch 216 to receive a radiation therapy dose, according to a radiation therapy treatment plan. The radiation therapy output 204 can be mounted or attached to a gantry 206 or other mechanical support. One or more chassis motors (not shown) may rotate the gantry 206 and the radiation therapy output 204 around couch 216 when the couch 216 is inserted into the treatment area. In an embodiment, gantry 206 may be continuously rotatable around couch 216 when the couch 216 is inserted into the treatment area. In another embodiment, gantry 206 may rotate to a predetermined position when the couch 216 is inserted into the treatment area. For example, the gantry 206 can be configured to rotate the therapy output 204 around an axis (“A”). Both the couch 216 and the radiation therapy output 204 can be independently moveable to other positions around the patient, such as moveable in transverse direction (“ ”), moveable in a lateral direction (“L”), or as rotation about one or more other axes, such as rotation about a transverse axis (indicated as “R”). A controller communicatively connected to one or more actuators (not shown) may control the couch 216 movements or rotations in order to properly position the patient in or out of the radiation beam 208 according to a radiation therapy treatment plan. Both the couch 216 and the gantry 206 are independently moveable from one another in multiple degrees of freedom, which allows the patient to be positioned such that the radiation beam 208 can target the tumor precisely. The MLC may be integrated and included within gantry 206 to deliver the radiation beam 208 of a certain shape.

The coordinate system (including axes A, T, and L) shown in FIG. 2A can have an origin located at an isocenter 210. The isocenter can be defined as a location where the central axis of the radiation beam 208 intersects the origin of a coordinate axis, such as to deliver a prescribed radiation dose to a location on or within a patient. Alternatively, the isocenter 210 can be defined as a location where the central axis of the radiation beam 208 intersects the patient for various rotational positions of the radiation therapy output 204 as positioned by the gantry 206 around the axis A As discussed herein, the gantry angle corresponds to the position of gantry 206 relative to axis A, although any other axis or combination of axes can be referenced and used to determine the gantry angle.

Gantry 206 may also have an attached imaging detector 214. The imaging detector 214 is preferably located opposite to the radiation source, and in an embodiment, the imaging detector 214 can be located within a field of the radiation beam 208.

The imaging detector 214 can be mounted on the gantry 206 (preferably opposite the radiation therapy output 204), such as to maintain alignment with the therapy beam 208. The imaging detector 214 rotates about the rotational axis as the gantry 206 rotates. In an embodiment, the imaging detector 214 can be a flat panel detector (e.g., a direct detector or a scintillator detector). In this manner, the imaging detector 214 can be used to monitor the radiation beam 208 or the imaging detector 214 can be used for imaging the patient's anatomy, such as portal imaging. The control circuitry of the radiation therapy device 202 may be integrated within the system 100 or remote from it.

In an illustrative embodiment, one or more of the couch 216, the therapy output 204, or the gantry 206 can be automatically positioned, and the therapy output 204 can establish the radiation beam 208 according to a specified dose for a particular therapy delivery instance. A sequence of therapy deliveries can be specified according to a radiation therapy treatment plan, such as using one or more different orientations or locations of the gantry 206, couch 216, or therapy output 204. The therapy deliveries can occur sequentially, but can intersect in a desired therapy locus on or within the patient, such as at the isocenter 210. A prescribed cumulative dose of radiation therapy can thereby be delivered to the therapy locus while damage to tissue near the therapy locus can be reduced or avoided.

FIG. 2B illustrates an exemplary radiation therapy device 202 that may include a combined linac and an imaging system, such as a CT imaging system. The radiation therapy device 202 can include an MLC (not shown). The CT imaging system can include an imaging X-ray source 218, such as providing X-ray energy in a kiloelectron-Volt (keV) energy range. The imaging X-ray source 218 can provide a fan-shaped and/or a conical radiation beam 208 directed to an imaging detector 222, such as a flat panel detector. The radiation therapy device 202 can be similar to the system described in relation to FIG. 2A, such as including a radiation therapy output 204, a gantry 206, a couch 216, and another imaging detector 214 (such as a flat panel detector) The X-ray source 218 can provide a comparatively-lower-energy X-ray diagnostic beam, for imaging.

In the illustrative embodiment of FIG. 2B, the radiation therapy output 204 and the X-ray source 218 can be mounted on the same rotating gantry 206, rotationally separated from each other by 90 degrees. In another embodiment, two or more X-ray sources can be mounted along the circumference of the gantry 206, such as each having its own detector arrangement to provide multiple angles of diagnostic imaging concurrently. Similarly, multiple radiation therapy outputs 204 can be provided.

FIG. 3 depicts an exemplary radiation therapy system 300 that can include combining a radiation therapy device 202 and an imaging system, such as a magnetic resonance (MR) imaging system (e.g., known in the art as an MR-linac) consistent with the disclosed embodiments. As shown, system 300 may include a couch 216, an image acquisition device 320, and a radiation delivery device 330. System 300 delivers radiation therapy to a patient in accordance with a radiotherapy treatment plan. In some embodiments, image acquisition device 320 may correspond to image acquisition device 132 in FIG. 1 that may acquire origin images of a first modality (e.g., MRI image shown in FIG. 4A) or destination images of a second modality (e.g., CT image shown in FIG. 4B).

Couch 216 may support a patient (not shown) during a treatment session. In some implementations, couch 216 may move along a horizontal translation axis (labelled “1”), such that couch 216 can move the patient resting on couch 216 into and/or out of system 300. Couch 216 may also rotate around a central vertical axis of rotation, transverse to the translation axis. To allow such movement or rotation, couch 216 may have motors (not shown) enabling the couch 216 to move in various directions and to rotate along various axes. A controller (not shown) may control these movements or rotations in order to properly position the patient according to a treatment plan.

In some embodiments, image acquisition device 320 may include an MRI machine used to acquire 2D or 3D MRI images of the patient before, during, and/or after a treatment session. Image acquisition device 320 may include a magnet 321 for generating a primary magnetic field for magnetic resonance imaging. The magnetic field lines generated by operation of magnet 321 may run substantially parallel to the central translation axis I. Magnet 321 may include one or more coils with an axis that runs parallel to the translation axis I. In some embodiments, the one or more coils in magnet 321 may be spaced such that a central window 323 of magnet 321 is free of coils. In other embodiments, the coils in magnet 321 may be thin enough or of a reduced density such that they are substantially transparent to radiation of the wavelength generated by radiotherapy device 330. Image acquisition device 320 may also include one or more shielding coils, which may generate a magnetic field outside magnet 321 of approximately equal magnitude and opposite polarity in order to cancel or reduce any magnetic field outside of magnet 321. As described below, radiation source 331 of radiation delivery device 330 may be positioned in the region where the magnetic field is cancelled, at least to a first order, or reduced.

Image acquisition device 320 may also include two gradient coils 325 and 326, which may generate a gradient magnetic field that is superposed on the primary magnetic field. Coils 325 and 326 may generate a gradient in the resultant magnetic field that allows spatial encoding of the protons so that their position can be determined. Gradient coils 325 and 326 may be positioned around a common central axis with the magnet 321 and may be displaced along that central axis. The displacement may create a gap, or window, between coils 325 and 326. In embodiments where magnet 321 can also include a central window 323 between coils, the two windows may be aligned with each other.

In some embodiments, image acquisition device 320 may be an imaging device other than an MRI, such as an X-ray, a CT, a CBCT, a spiral CT, a PET, a SPECT, an optical tomography, a fluorescence imaging, ultrasound imaging, radiotherapy portal imaging device, or the like. As would be recognized by one of ordinary skill in the art, the above description of image acquisition device 320 concerns certain embodiments and is not intended to be limiting.

Radiation delivery device 330 may include the radiation source 331, such as an X-ray source or a linac, and an MLC 332 (shown below in more detail in FIG. 5). Radiation delivery device 330 may be mounted on a chassis 335. One or more chassis motors (not shown) may rotate the chassis 335 around the couch 216 when the couch 216 is inserted into the treatment area. In an embodiment, the chassis 335 may be continuously rotatable around the couch 216, when the couch 216 is inserted into the treatment area. Chassis 335 may also have an attached radiation detector (not shown), preferably located opposite to radiation source 331 and with the rotational axis of the chassis 335 positioned between the radiation source 331 and the detector. Further, the device 330 may include control circuitry (not shown) used to control, for example, one or more of the couch 216, image acquisition device 320, and radiotherapy device 330. The control circuitry of the radiation delivery device 330 may be integrated within the system 300 or remote from it.

During a radiotherapy treatment session, a patient may be positioned on couch 216. System 300 may then move couch 216 into the treatment area defined by the magnet 321, coils 325, 326, and chassis 335. Control circuitry may then control radiation source 331, MLC 332, and the chassis motor(s) to deliver radiation to the patient through the window between coils 325 and 326 according to a radiotherapy treatment plan.

FIG. 2A, FIG. 2B, and FIG. 3 illustrate generally embodiments of a radiation therapy device configured to provide radiotherapy treatment to a patient, including a configuration where a radiation therapy output can be rotated around a central axis (e.g., an axis “A”). Other radiation therapy output configurations can be used. For example, a radiation therapy output can be mounted to a robotic arm or manipulator having multiple degrees of freedom. In yet another embodiment, the therapy output can be fixed, such as located in a region laterally separated from the patient, and a platform supporting the patient can be used to align a radiation therapy isocenter with a specified target locus within the patient.

As discussed above, radiation therapy devices described by FIG. 2A, FIG. 2B, and FIG. 3 include an MLC for shaping, directing, or modulating an intensity of a radiation therapy beam to the specified target locus within the patient. FIG. 5 illustrates an exemplary MLC 332 that includes leaves 532A through 532J that can be automatically positioned to define an aperture approximating a tumor 540 cross-section or projection. The leaves 532A through 532J permit modulation of the radiation therapy beam. The leaves 532A through 532J can be made of a material specified to attenuate or block the radiation beam in regions other than the aperture, in accordance with the radiation treatment plan. For example, the leaves 532A through 532J can include metallic plates, such as comprising tungsten, with a long axis of the plates oriented parallel to a beam direction and having ends oriented orthogonally to the beam direction (as shown in the plane of the illustration of FIG. 2A). A “state” of the MLC 332 can be adjusted adaptively during a course of radiation therapy treatment, such as to establish a therapy beam that better approximates a shape or location of the tumor 540 or other target locus. This is in comparison to using a static collimator configuration or as compared to using an MLC 332 configuration determined exclusively using an “offline” therapy planning technique. A radiation therapy technique using the MLC 332 to produce a specified radiation dose distribution to a tumor or to specific areas within a tumor can be referred to as IMRT. The resulting beam shape that is output using the MLC 332 is represented as a graphical aperture image. Namely, a given graphical aperture image is generated to represent how a beam looks (beam shape) and its intensity after being passed through and output by MLC 332.

FIG. 6 illustrates an embodiment of another type of radiotherapy device 630 (e.g., a Leksell Gamma Knife), according to some embodiments of the present disclosure. As shown in FIG. 6, in a radiotherapy treatment session, a patient 602 may wear a coordinate frame 620 to keep stable the patient's body part (e.g., the head) undergoing surgery or radiotherapy. Coordinate frame 620 and a patient positioning system 622 may establish a spatial coordinate system, which may be used while imaging a patient or during radiation surgery. Radiotherapy device 630 may include a protective housing 614 to enclose a plurality of radiation sources 612. Radiation sources 612 may generate a plurality of radiation beams (e.g., beamlets) through beam channels 616. The plurality of radiation beams may be configured to focus on an isocenter 210 from different directions. While each individual radiation beam may have a relatively low intensity, isocenter 210 may receive a relatively high level of radiation when multiple doses from different radiation beams accumulate at isocenter 210. In certain embodiments, isocenter 210 may correspond to a target under surgery or treatment, such as a tumor.

FIG. 7A illustrates an exemplary data flow for training and use of a GAN adapted for generating a synthetic graphical aperture image representation of multi-leaf collimator (MLC) leaf positions from a received projection image that represents a view of an anatomy of a subject image. For instance, the generator model 732 of FIG. 7A, which is trained to produce a trained generator model 760, may be trained to implement the processing functionality provided as part of the image processor 114 in the radiotherapy system 100 of FIG. 1. Accordingly, a data flow of the GAN model usage 750 (prediction) is depicted in FIG. 7A as the provision of new data 770 (e.g., projection image that represents a view of an anatomy of a subject input images from a new patient) to a trained generator model 760, and the use of the trained generator model 760 to produce a prediction or estimate of a generator output (images) 734 (e.g., synthetic graphical aperture images corresponding to the input projection image that represents a view of an anatomy of a subject image). A projection image can be generated from one or more CT or MR images of a patient anatomy representing a view of the anatomy from a given angle of the gantry.

GANs comprise two networks: a generative network (e.g., generator model 732) that is trained to perform classification or regression, and a discriminative network (e.g., discriminator model 740) that samples the generative network's output distribution (e.g., generator output (images) 734) or a training graphical aperture image from the training images 723 and decides whether that sample is the same or different from the true test distribution. The goal for this system of networks is to drive the generator network to learn the ground truth model as accurately as possible such that the discriminator net can only determine the correct origin for generator samples with 50% chance, which reaches an equilibrium with the generator network. The discriminator can access the ground truth but the generator only accesses the training data through the response of the detector to the generator's output.

The data flow of FIG. 7A illustrates the receipt of training input 710, including various values of model parameters 712 and training data 720 (with such training images 723 including a set of projection images that represent different views of an anatomy of subject patient imaging data paired with real graphical aperture image representations of multi-leaf collimator (MLC) leaf positions corresponding to the patient imaging data at the different views, and conditions or constraints 726. The training input 710 is provided to the GAN model training 730 to produce a trained generator model 760 used in the GAN model usage 750.

As part of the GAN model training 730, the generator model 732 is trained on real training graphical aperture images and corresponding training projection images that represent views of an anatomy of a subject image pairs 722 (also depicted in FIG. 7B as 723), to produce and map segment pairs in the CNN. In this fashion, the generator model 732 is trained to produce, as generator output (images) 734, simulated or synthetic graphical aperture images of MLC leaf positions based on an input map. The discriminator model 740 decides whether a simulated graphical aperture images is from the training data (e.g., the training or true graphical aperture image representations of MLC leaf positions) or from the generator (e.g., the synthetic graphical aperture image representations, as communicated between the generator model 732 and the discriminator model 740). The discriminator output 736 is a decision of the discriminator model 740 indicating whether the received image is a simulated image or a true image and is used to train the generator model 732. In some cases, the generator model 732 is trained utilizing the discriminator on the generated images and is further trained based on cycle-consistency loss information. This training process results in back-propagation of weight adjustments 738, 742 to improve the generator model 732 and the discriminator model 740.

During training of generator model 732, a batch of training data can be selected from the patient images and expected results. The selected training data can include at least one projection image of patient anatomy representing a view of the patient anatomy from a given gantry angle and the corresponding training or real graphical aperture image and/or machine parameter data at that given gantry angle. The selected training data can include multiple projection images of patient anatomy representing views of the same patient anatomy from multiple equally spaced or non-equally spaced gantry angles (e.g., from 0 degrees, from 15 degrees, from 45 degrees, from 60 degrees, from 75 degrees, from 90 degrees, from 105 degrees, from 120 degrees, from 135 degrees, from 150 degrees, from 165 degrees, from 180 degrees, from 195 degrees, from 210 degrees, from 225 degrees, from 240 degrees, from 255 degrees, from 270 degrees, from 285 degrees, from 300 degrees, from 315 degrees, from 330 degrees, from 345 degrees, and/or from 360 degrees) and the corresponding training graphical aperture image and/or machine parameter data at those different equally-spaced or non-equally spaced gantry angles.

Thus, in this example, data preparation for the GAN model training 730 requires graphical aperture image representation of MLC leaf positions that are paired with projection images that represent views of an anatomy of subject images (these may be referred to as training projection images that represent a view of an anatomy of a subject image at various gantry angles). Namely, the training data includes paired sets of graphical aperture images at the same gantry angles as the corresponding projection images. In an example, the original data includes pairs of projection images that represents a view of an anatomy of a subject at various gantry angles and corresponding graphical aperture image representation of MLC leaf positions at the corresponding gantry angles that may be registered and resampled to a common coordinate frame to produce pairs of anatomy-derived images. The training data can include multiple of these paired images for multiple patients at any number of different gantry angles. In some cases, the training data can include 360 pairs of projection images and graphical aperture images, one for each angle of the gantry for each training patient. In some cases, 8 pairs may be included where each pair represents a different 45-degree view of the anatomy relative to the gantry.

The expected results can include estimated or synthetic graphical aperture image representation of leaves of the MLC that can be used to compute control points for generating a beam shape at the corresponding gantry angle to define the delivery of radiation treatment to a patient. The control points or machine parameters can include at least one gantry angle, at least one multi-leaf collimator leaf position, and at least one aperture weight or intensity.

In detail, in a GAN model, the generator (e.g., generator model 732) learns a distribution over the data x, p_G(x), starting with noise input with distribution p_z(z) as the generator learns a mapping G(z; θ_G): p_Z(z)→p^G(x) where G is a differentiable function representing a neural network with layer weight and bias parameters θ_G. The discriminator, D(x; θ_D) (e.g., discriminator model 740), maps the generator output to a binary scalar (true, false), deciding true if the generator output is from actual data distribution p_data(x) and false if from the generator distribution p_G(x). That is, D(x) is the probability that x came from p_data(x) rather than from p_G(x). In another embodiment, paired training data may be utilized in which, for instance, Y is conditioned (dependent) on X. In such cases, the GAN generator mapping is represented by G(y|x; θ_G): X→Y from data domain X where data xϵX represents the anatomy projection images and domain Y where data yϵY represents the control point apertures corresponding to x. Here an estimate for an aperture is conditioned on its projection. Another difference from the straight GAN is that instead of a random noise z input, the projection image x is the generator input. For this embodiment the setup of the discriminator is the same as above. In general, the generator model 732 and the discriminator model 740 are in a circular data flow, where the results of one feed into the other. The discriminator takes either training or generated images and its output is used to both adjust the discriminator weights and to guide the training of the generator network.

FIG. 7B illustrates an exemplary CNN model adapted for generating a synthetic graphical aperture image representation of the MLC leaf positions at one or more gantry angles (or graphical aperture image) according to the present disclosure. Specifically, the model shown in FIG. 7B depicts an arrangement of a “U-Net” deep CNN designed for generating an output data set (output graphical aperture images 736) based on an input training set (e.g., paired projection images and graphical aperture images 723). The name derives from the “U” configuration, and as is well understood, this form of CNN model can produce pixel-wise classification or regression results. In some cases, a first path leading to the CNN model includes one or more deformable offset layers and one or more convolution layers including convolution, batch normalization, and an activation such as the rectified linear unit (ReLU) or one of its variants. The model generates as output data set graphical aperture images 736.

The left side of the model operations (the “encoding” operations 792) learns a set of features that the right side (the “decoding” operations 794) uses to reconstruct an output result. The U-Net has n levels consisting of conv/BN/ReLU (convolution/batch normalization/rectified linear units) blocks 790, and each block has a skip connection to implement residual learning. The block sizes are denoted in FIG. 7B by “S” and “F” numbers; input images are SxS in size, and the number of feature layers is equal to F. The output of each block is a pattern of feature responses in arrays the same size as the images.

Proceeding down the encoding path, the size of the blocks decreases by ½ or 2⁻¹ at each level while the size of the features by convention increases by a factor of 2. The decoding side of the network goes back up in scale from S/2ⁿ while adding in feature content from the left side at the same level; this is the copy/concatenate data communication. The differences between the output image and the training version of that image drives the generator network weight adjustments by backpropagation. For inference, or testing, with use of the model, the input would be a single image or collection of images of the projection images 723 (e.g., at different gantry angles) and the output would be graphical aperture images 736 (e.g., graphical images corresponding to the different gantry angles).

The representation of the model of FIG. 7B thus illustrates the training and prediction of a generative model, which is adapted to perform regression rather than classification. FIG. 7C illustrates an exemplary CNN model adapted for discriminating a synthetic graphical aperture image(s) according to the present disclosure. The discriminator network shown in FIG. 7C may include several levels of blocks configured with stride-2 convolutional layers, batch normalization layers and ReLU layers, and separated pooling layers. At the end of the network, there will be one or a few fully connection layers to form a 2D patch for discrimination purposes. The discriminator shown in FIG. 7C may be a patch-based discriminator configured to receive input graphical aperture images (e.g., generated from the generator shown in FIG. 7B), classify the image as real or fake, and provide the classification as output detection results 744.

Consistent with embodiments of the present disclosure, the treatment modeling methods, systems, devices, and/or processes based on such models include two stages: training of the generative model, with use of a discriminator/generator pair in a GAN; and prediction with the generative model, with use of a GAN-trained generator. Various examples involving a GAN and a CycleGAN for graphical aperture images generation are discussed in detail in the following examples. It will be understood that other variations and combinations of the type of deep learning model and other neural-network processing approaches may also be implemented with the present techniques. Further, although the following examples are discussed with reference to images and image data, it will be understood that the following networks and GAN may operate with use of other non-image data representations and formats.

In deep CNN training, the learned model is the values of layer node parameters θ (node weights and layer biases) determined during training. Training employs maximum likelihood or the cross entropy between the training data and the model distribution. A cost function expressing this relationship is

(θ)=−E_x,y˜pdata log p_model(y|x;θ)

The exact form of the cost function for a specific problem depends on the nature of the model used. A Gaussian model p_model(y|x)=N(y: f(x; θ)) implies a cost function such as:

(θ)=−E_x,y˜pdata∥y−f(x;θ)∥₂²+const

Which includes a constant term that does not depend on θ. Thus, minimizing (θ) generates the mapping f(x; θ) that approximates the training data distribution.

A useful extension of the GAN is the conditional GAN. Conditional adversarial networks learn a mapping from observed image x and random noise z as

$G\text{:}\left\{x,z\right\}\to y.$

Both adversarial networks consist of two networks: a discriminator (D) and a generator (G). The generator G is trained to produce outputs that cannot be distinguished from “real” or actual training images by an adversarially trained discriminator D that is trained to be maximally accurate at detecting “fakes” or outputs of G. The conditional GAN differs from the unconditional GAN in that both discriminator and generator inferences are conditioned on an example image of the type X.

The conditional GAN loss function is:

${ℒ}_{\mathrm{cGAN}}\left(G,D\right)={𝔼}_{x,y\sim p_{\mathrm{data}}\left(x,y\right)}\left[\log \left(D\left(x,y\right)\right)\right]+E_{x\sim p_{\mathrm{data}}\left(x\right),z\sim p_z\left(z\right)}\left[\log \left(1-D\left(x,G\left(x,z\right)\right)\right)\right]$

where G tries to minimize this loss against an adversarial D that tries to maximize it, or

$G^{*}=\underset{G}{\mathrm{argmin}}\underset{D}{\max }{ℒ}_{\mathrm{cGAN}}\left(G,D\right).$

In addition, one wants the generator G to minimize the difference between the training estimates and the actual training ground truth images, _L1(G)=_{x,y˜pdata(x,y),z˜pz(z)}[∥y−G(x,z)∥₁] so the complete loss is the λ-weighted sum of two losses

$G^{*}=\underset{G}{\mathrm{argmin}}\underset{D}{\max }{ℒ}_{\mathrm{cGAN}}\left(G,D\right)+\lambda {ℒ}_{L1}\left(G\right).$

The generator in the conditional GAN may be a U-Net.

According to some embodiments, the generator of the conditional GAN is trained to receive a projection image and generates a synthetic graphical aperture image corresponding to the projection image. The discriminator receives the synthetic graphical aperture image from the generator and is trained to distinguish the received synthetic graphical aperture image from being a real image or a fake or synthetic image. The generator is trained to minimize a difference between the synthetic graphical aperture image and the corresponding training graphical aperture image. To this end, after the generator generates the synthetic graphical aperture image, the training graphical aperture image corresponding to the projection image is retrieved. For example, the projection image may be an image representing a 2D or 3D projection of a CT image from a gantry angle of 15 degrees during a specific time interval of a treatment fraction for a given patient that was previously treated. A graphical aperture image may be generated representing the radiotherapy treatment machine parameters for the very same patient, same specific time interval and the same gantry angle of 15 degrees. This graphical aperture image is paired with the projection image and is retrieved as a training graphical aperture image. A comparison is made between the generated synthetic graphical aperture image for the projection image and this retrieved training graphical aperture image. Parameters of the generator are then updated based on the difference in an attempt to minimize the difference. The discriminator is similarly trained based on whether the discriminator correctly classifies the generated image as real or fake. The parameters of the generator are updated based on a total loss function that takes into account the discriminator errors and the generator differences.

Subsequently, a second projection image for the same or different patient and at the same or different gantry angle is retrieved together with the paired training graphical aperture image. The generator receives this second projection image and generates a second synthetic graphical aperture image. A difference between the paired training graphical aperture image and the synthetically generated graphical aperture image is made and parameters of the generator are again updated based on this difference. Once all or a substantial portion of the training data is processed and used to update parameters of the generator and/or once a specified number of epochs or when the error (between synthetic graphical aperture images and training graphical aperture images) is within a threshold, the training ends and the generator's parameters are output.

Another useful extension of a GAN is the CycleGAN, which is described below in connection with FIG. 7D. FIG. 7D illustrates training and use of CycleGAN 731 for generating a collection of graphical aperture images (e.g., a collection of synthetic or estimated graphical aperture images) from a received collection of projection images according to some examples of the disclosure. CycleGAN 731 includes a first generator model 735, second generator model 737, a first discriminator model 739A, and a second discriminator model 739B. The first generator model 735 includes deformable offset layers and convolution blocks and the second generator model 737 includes deformable offset layers and convolution blocks. These two models 735 and 737 may each be an implementation of generator model 732 (FIG. 7A) (e.g., as regression-type DCNN models), and first discriminator model 739A and second discriminator model 739B may each be an implementation of discriminator model 740 (e.g., as classification-type DCNN models). CycleGAN 731 may be divided into two portions, first portion 733A and second portion 733B.

The convolution blocks of each generator model 735 and 737 may be trained together or separate from training of the other generator and discriminator models. Specifically, the convolution blocks of the generator models 735 and 737 are trained to obtain the correct weights to perform their function. The deformable offset layers may each be trained to coordinate offsets, resample, and perform interpolation. The deformable offset layers may be trained together or separate from training of the generator and discriminator models. The effect of these offset layers changes the original regular sampling grids from upper convolutional blocks, introduces coordinate offsets, and resamples the images using interpolation. The deformable offset layers may alternatively or in addition be implemented using a spatial transformer, other types of convolutional layers, and/or any other module that can store deformed structure information for an image. The number of offset layers in the deformable offset layers may vary based on image size, the number of down-sampling convolutional layers, and other factors.

In an example, in first portion 733A, the first generator model 735 may be trained to receive a training collection of projection images 723A (which may include one of image pairs 722) and generates a respective synthetic first collection of graphical aperture images as first generation results 736A. The first generator model 735 is referred to as G^{proj2aperture}.

First generation results 736A may be provided to the first discriminator model 739A. The first discriminator model 739A may classify the synthetic collection of graphical aperture images as a real collection of graphical aperture training images or a simulated collection of graphical aperture training images and provide the classification as detection results 744A. The first generation results 736A and detection results 744A may be fed back to the first generator model 735 and first discriminator model 739A to adjust weights implemented by the first generator model 735 and first discriminator model 739A. For example, first generation result 736A (e.g., a collection of graphical aperture images generated by first generator model 735) and detection results 744A may be used to calculate adversarial losses.

The first generation results 736A (e.g., the synthetic collection of graphical aperture images) may also be concurrently provided to the second generator model 737. The second generator model 737 may receive first generation results 736A and generate a respective simulated collection of projection images as outputs. The simulated collection of projection images may be referred to as cycle collection of projection images 741 and may be used to compute cycle losses to adjust weights of first/second generator model 735/737. The second generator model 737 that generates the first cycle collection of projection images 741 is referred to as G^{aperture2proj}.

In an example, in the second portion 733B, the second generator model 737 may be trained to receive a real collection of training graphical aperture images 723B (which may include one of image pairs 722) and to generate a respective synthetic collection of projection images (synthetic or simulated collection of projection images) as first generation results 736B. The second generator model 737 that generates the first generation results 736B is the same generator as that used in the first portion 733A.

First generation results 736B may be provided to the second discriminator model 739B. The second discriminator model 739B may classify the synthetic collection of projection images as a real collection of projection training images or a simulated collection of projection training images and provide the classification as detection results 744B. The first generation results 736B and detection results 744B may be fed back to the second generator model 737 and the second discriminator model 739B to adjust weights implemented by second generator model 737 and second discriminator model 739B. For example, first generation result 736B (e.g., a synthetic collection of projection images generated by second generator model 737) and the detection results 744B may be used to calculate adversarial losses.

First generation results 736B (e.g., synthetic collection of projection images) may also be concurrently provided to the first generator model 735. The first generator model 735 may receive first generation results 736B and generate respective cycle-graphical aperture images 743 as outputs. The cycle-graphical aperture images 743 may be used to compute cycle losses to adjust weights of first/second generator model 735/737. The first generator model 735 that generates the cycle-graphical aperture images 743 is the same generator as that used in the first portion 733A, and the second generator model 737 that generates the cycle-graphical aperture images 743 is the same generator as that used in the first portion 733A.

In some examples, “adversarial losses” may account for the classification losses for the first and second discriminator models 739A and 739B. First and second discriminator models 739A and 739B may classify whether the synthetic images have similar distribution as true images or not. For cycle-consistency losses, the losses are calculated between each pair of true collection of projection images and cycle-collection of projection images, and each pair of true graphical aperture images and cycle-graphical aperture image, respectively. For example, a first loss may be calculated between a collection of projection training images 723A and cycle-collection of projection images 741 and a second loss may be calculated between real training collection of graphical aperture images 723B and collection of cycle-graphical aperture images 743. The cycle-collection of projection images 741 and cycle-graphical aperture images 743 may both be obtained by doing forward and backward cycles. Each pair of true collection of projection images 723A and cycle-collection of projection images 741 may be in the same collection of projection images domain, and each pair of real training graphical aperture images 723B and cycle-graphical aperture images 743 may be in the same graphical aperture image domain. The CycleGAN 731 may accordingly rely on a whole pool (or a plurality) of true or real projection training images 723A and a whole pool (or a plurality) of real training graphical aperture images 723B to produce synthetic graphical aperture images (collection of graphical aperture images), synthetic collection of projection images, cycle-collection of projection images 741, and cycle-graphical aperture images 743. Based on “adversarial losses” and “cycle-consistency losses,” CycleGAN 731 may produce sharp synthetic graphical aperture images, which have similar image resolution as real graphical aperture images.

In some examples, a processor (e.g., of system 100) may apply image registration to register real graphical aperture training images to training collection of projection images. This may create a one-to-one corresponding relationship between projection images at different gantry angles and graphical aperture images at each of the different gantry angles in the training data. This relationship may be referred to as paired or a pair of projection images and graphical aperture images.

In some implementations, CycleGAN 731 may be implemented to generate a collection of graphical aperture images in accordance with an objective function that includes an adversarial loss term and a cycle consistency loss term. The cycleGAN network has two separate adversarial losses. Similar to the conditional GAN, the mapping G:X→Y and its associated discriminator D_Y gives loss represented by _GAN(G, D_Y)=_y˜pdata(x)[log(1−D_Y(G(x)))].

The adversarial losses may be determined using first/second generator models 735/737 and first/second discriminator models 739A/739B and may be expressed as:

Adversarial Losses:

L _GANs =L _GAN(G^pro2aperture,D_aperture,x,y)+L_GAN(G^pro2aperture,D_proj,x,y)

where:

L _GAN(G^pro2aperture,D_aperture,x,y)=E_x-p(proj)[D_ct(G^{proj2aperture}(x))]²+E_{y-p(aperture)}[D_aperture(y)−1]²

L _GAN(G^{aperture2proj},D_proj,x,y)=E_{y-p(aperture)}[D_proj(G^{aperture2proj}(y))]²+E_x-p(proj)[D_proj(x)−1]²

D_aperture is the first discriminator model which determines whether one image is a true collection of graphical aperture images or a synthetic collection of graphical aperture images. D_proj is the second discriminator model which determines whether one image is a true collection of projection images or a synthetic collection of projection images.

Cycle-consistence losses are applied on both generators G^{proj2aperture} and G^{aperture2proj}. Such cycle consistency loss term may be determined using images 741 and 743 and may be expressed as:

$L_{\mathrm{CYC}}\left(G^{\mathrm{proj}2\mathrm{aperture}},G^{\mathrm{aperture}2\mathrm{proj}},x,y\right)=E_{x\sim p\left(\mathrm{proj}\right)}{G^{\mathrm{aperture}2\mathrm{proj}}\left(G^{\mathrm{proj}2\mathrm{aperture}}\left(x\right)\right)-x}_1+E_{y\sim p\left(\mathrm{aperture}\right)}{G^{\mathrm{proj}2\mathrm{aperture}}\left(G^{\mathrm{aperture}2\mathrm{proj}}\left(y\right)\right)-y}_1$

The effect of minimizing the “cycle-consistence” loss terms is to preserve original structures and to avoid unnecessary structure deformation, and the effect of minimizing the “adversarial” loss terms is to learn a mapping or distribution conversion from one domain to its opponent's domain. An identity loss can be used to regularize the generator to be near an identity mapping when real samples of the target domain Y are input. The identity loss can be represented by _identity(G, F)=_y˜pdata(y)[∥G(y)−y|∥₁]+_x˜pdata(x)[∥F(x)−x∥₁].

Accordingly, the total objective function can be expressed as: Total loss: L_total=L_GANS+Δ_cyc·L_CYC+λ_identity·L_identity, where λ_cycle and λ_identity control the relative strength of the losses or as: (G, f, D_X, D_Y)=_GAN(G, D_Y)+_GAN(F, D_X)+Δ_cyccyc(G, F)+Δ_{identityidentity}(G, F).

CycleGAN 731 may train first and second generators 735/737 in accordance with:

$G^{\mathrm{proj}2\mathrm{aperture}},G^{\mathrm{aperture}2\mathrm{proj}}=\hspace{1em}\hspace{1em}\arg \underset{G^{\mathrm{proj}2\mathrm{aperture}},G^{\mathrm{aperture}2\mathrm{proj}}}{\min }\underset{D_{\mathrm{aperture}},D_{\mathrm{proj}}}{\max }{L}_{\mathrm{total}}\hspace{1em}\left(G^{\mathrm{proj}2\mathrm{aperture}},G^{\mathrm{aperture}2\mathrm{proj}},D_{\mathrm{aperture}},D_{\mathrm{proj}}\right),$

which can be done by some common optimization algorithms used in the deep learning field, such as stochastic gradient descent, Adam's method, or other popular methods.

The preceding examples provide an example of how a GAN, a conditional GAN, or CycleGAN may be trained based on a collection of graphical aperture images and collection of projection images pairs, specifically from image data in 2D or 3D image slices in multiple parallel or sequential paths. It will be understood that the GAN, conditional GAN, or CycleGAN may process other forms of image data (e.g., 3D, or other multi-dimensional images). Further, although only grayscale (including black and white) images are depicted by the accompanying drawings, it will be understood that color images may be generated and/or processed by the GAN, as discussed in the examples below.

As an example, the CycleGAN can be trained using a batch of training data that includes paired projection images and graphical aperture images for various patients. Specifically, a first batch of training data can include a collection of projection images for a first patient that were used to generate a treatment plan for treating the first patient in a first treatment fraction. Based on the treatment plan, a corresponding collection of graphical aperture images can be generated. Each graphical aperture image is associated with the first patient and is associated with the first treatment fraction. Each graphical aperture image is paired with the projection images such that the projection image for a given gantry angle at a given point in time in the treatment fraction is associated with a corresponding graphical aperture image.

In some implementations, the first batch of training data is processed by the CycleGAN to train the first and second generators and the first and second discriminators. Specifically, the first generator in the first portion of the CycleGAN receives a first projection image corresponding to a first gantry angle and a first portion of the treatment fraction for the first patient. The first generator generates a first synthetic graphical aperture image corresponding to the first projection image. The first synthetic graphical aperture image is provided to the first discriminator, which attempts to distinguish whether the first synthetic graphical aperture image is a real or training graphical aperture image or a synthetic graphical aperture image. The first training graphical aperture image, corresponding to the first projection image, for the first gantry angle and the first portion of the treatment fraction for the first patient is retrieved and used in the loss function to train the first generator and the first discriminator.

The first synthetic graphical aperture image is concurrently or sequentially provided to the second generator. The second generator generates a cycle projection image corresponding to the first synthetic graphical aperture. The cycle projection image is then compared with the first projection image to generate a cycle loss. The first generator, the second generator, and the first discriminator are trained based on the loss function that includes the cycle loss.

In parallel with the first portion receiving the first projection image, the second generator in the second portion of the CycleGAN receives the first training graphical aperture image, corresponding to the first projection image. The second generator generates a first synthetic projection image corresponding to the first training graphical aperture image. The first synthetic projection image is provided to the second discriminator, which attempts to distinguish whether the first synthetic projection image is a real or training projection image or a synthetic projection image. The first projection image, corresponding to the first training graphical aperture image, for the first gantry angle and the first portion of the treatment fraction for the first patient is retrieved and used in the loss function to train the second generator and the second discriminator.

The first synthetic projection image is concurrently or sequentially provided to the first generator. The first generator generates a cycle graphical aperture image corresponding to the first synthetic projection image. The cycle graphical aperture image is then compared with the first training graphical aperture image to generate a cycle loss. The first generator, the second generator, and the second discriminator are trained based on the loss function that includes the cycle loss.

The first and second portions then retrieve another batch of training data and perform similar operations for another paired set of projection and graphical aperture images for another patient for which a treatment plan was previously generated. Once a specified number of epochs or training data is processed and/or when the loss function reaches a specified threshold value, training completes and the first generator is output and used to generate graphical aperture images for a new collection of projection images corresponding to a new patient or subject.

FIG. 10 illustrates an embodiment of a method 1000 for training a GAN, such as the CycleGAN, for determining a set of machine parameters or graphical aperture image(s) for that particular gantry angle based on at least one medical image projection depicting a view of a patient anatomy from a gantry angle. The GAN can receive sets of medical images representing different image projections depicting views of a patient anatomy from different gantry angles. The projection images can be generated from CT images, MRI images, synthetic CT, and/or PET images. The GAN can also receive corresponding anatomy voxel label maps and functions of labelled object signed distance maps. Machine parameters or graphical aperture image(s) for the various gantry angles depicted in the projection images can then be determined for each set of medical images. In an embodiment, the machine parameters or graphical aperture image(s) can be received together with the sets of medical projection images.

The machine parameters can include at least one gantry angle, at least one multi-leaf collimator leaf position, and at least one aperture weight or intensity. The machine parameters can be represented by the expression Y={ϕ, ( . . . L_n^ϕ, R_n^ϕ . . . )^T, y^ϕ}_k, k=1, . . . , K, where Y can represent a set of K machine parameter data objects, ϕ can represent a gantry angle, L_n^ϕ, R_n^ϕ can represent the n-th left and right multi-leaf collimator leaf positions at angle ϕ, and y^ϕ can represent an aperture weight or intensity for a gantry angle ϕ. To begin network training, an iteration index can be set to an initial value of zero. A batch of training data can be formed from a subset of the received sets of medical image projections and corresponding graphical aperture image(s) for the gantry angles represented by the medical image projections. Particularly, one batch of medical image projections may represent anatomy of a first training subject from multiple gantry angles and the training data includes graphical aperture images generated based on machine parameters that were used at each of the multiple gantry angles while treating the first training subject. Another batch of medical image projections may represent anatomy of a second training subject from multiple gantry angles and the training data includes graphical aperture images generated based on machine parameters that were used at each of the multiple gantry angles while treating the second training subject. [Swapped out the old math graphics for the new Word math graphics.]

The batch of training data can be provided to the GAN and the GAN parameters can be updated based thereon. The GAN can provide an output set of machine parameters or graphical aperture images based on current parameters of the GAN for a given set of received media image projections A comparison can be made between the output set of machine parameters or graphical aperture images corresponding to the received sets of medical image projections in the batch of training data (e.g., training machine parameters or graphical aperture images). Corresponding error sets, where each error value can be the difference between the estimated machine parameters or graphical aperture images and the corresponding training machine parameters or graphical aperture images are determined from the comparison. Parameters of the GAN can then be updated based on the corresponding errors, such as by using backpropagation.

In an embodiment, parameters of the GAN, Θ, can be updated, such as to minimize or reduce a cost function, such as the cost function (Θ*=argmin_Θ∥Y−Y*∥₂²+λ∥Θ∥₁, where Y can represent the machine parameters or graphical aperture images determined by the GAN, where Y* can represent the known machine parameters or graphical aperture images corresponding to the batch of training data, and where Θ* can represent parameters of the GAN (e.g., layer node weights and biases as described above) corresponding to a minimized square error between Y and Y*.

In an embodiment, the cost function can include a probabilistic function where parameters of the GAN can be determined according to the expression Θ_train=argmax_Θ(Y|X; Θ), or Θ_train=argmax_ΘΣ_tϵT log P(Y_t|X_t; Θ), where Θ_train can represent the parameters of the fully trained GAN, and X can represent a collection of medical image projections at different gantry angles, derived from anatomy voxel label maps, or functions of the labelled object signed distance maps.

After updating the parameters of the GAN, the iteration index can be incremented by a value of one. The iteration index can correspond to a number of times that the parameters of the GAN have been updated. Stopping criteria can be computed, and if the stopping criteria are satisfied, then the GAN model can be saved in a memory, such as the memory device 116 of image processing device 112, and the training can be halted. If the stopping criteria are not satisfied, then the training can continue by obtaining another batch of training images from the same training subject or another training subject. In an embodiment, the stopping criteria can include a value of the iteration index (e.g., the stopping criteria can include whether the iteration index is greater than or equal to a determined maximum number of iterations). In an embodiment, the stopping criteria can include an accuracy of the output set of machine parameters or graphical aperture images (e.g., the stopping criteria can include whether the difference between the output set of machine parameters or graphical aperture images and the machine parameters or graphical aperture images corresponding to the received sets of medical image projections in the batch of training data is smaller than a threshold). In an embodiment, the threshold can correspond to an asymptotic minimum of all errors determined. In an embodiment, the machine parameters or graphical aperture images can be presented to the GAN in the form of images with fixed formats specifying, for example, apertures, angles, and intensity values. In an embodiment, the patient images can be pooled with machine parameters and can be presented as real arrays.

A motivation for predicting machine parameters or control points, using a machine learning model, is to accelerate treatment planning computations. Current, conventional treatment planning proceeds from CT or MR image through anatomy and target delineation to planning. Modern IMRT and VMAT planning uses the images and the delineated structures to define a 3D dose distribution and the machine parameters to deliver that dose distribution. The resulting clinical plan based on the CT or MR image is typically produced after two phases of numerical optimization. In the case of an MR image, an additional step is provided to convert the MR image to a synthetic CT image. The CT image pixel intensities are a function of the tissues' X-ray absorption and provide essential physics information for the treatment planning program. MR-based planning creates a synthetic CT image to provide that X-ray absorption data for planning. The first phase (e.g., Fluence Map Optimization or “FMO”) produces an idealized dose distribution that satisfies the planner's requirements. The second phase (e.g., Segment Shape, Weight Optimization) creates the machine parameters for the given linac and MLC to produce a 3D dose distribution in the patient that as closely as possible agrees with the Fluence Map distribution.

The disclosed embodiments improve the quality and speed at which the treatment plans are created using control point prediction (e.g., a DL Plan Estimate). The GAN workflows imagine GAN segmentation of anatomy and target and the generation of synthetic CTs for MR images. GAN training produces a model for control points based on very good treatment plans involving both phases of optimization. The control points estimated by the trained GAN in a sense incorporate the information in both planning phases. One way in which the planning process can be shortened includes eliminating the Fluence Map phase altogether and using the control point estimates as input to the second phase (Segment Shape, Weight Optimization). This speeds up the second phase computation, which speeds up the planning process.

A challenge for control point prediction based on anatomy is that the anatomy and control points have fundamentally different common representations. Anatomies are depicted by rectilinear medical images of various modalities and control points are vectors of real number parameters. Further, even if the control points' apertures are represented by a graphical representation in an image, the orientation of the aperture does not correspond to any of the standard 2D or 3D views of anatomy. As the linac travels in an arc around the patient, the anatomy view at any moment is a projection image of the anatomy, equivalent to a plane radiograph of that anatomy at that angle. According to some embodiments, control point aperture data is re-formatted, converted to an aperture image, and aligned with the anatomy projections at the corresponding angles.

Once trained, the GAN can compute estimates of graphical aperture images 1010 for one or more gantry angles for a new patient (e.g., a patient that did not contribute known images or image functions used in the training process) given only the new patient's projection images 850 at the one or more gantry angles. In an embodiment, the GAN can include a GAN for regression. The GAN can be stored in a memory, such as the memory device 116 of image processing device 112. The GAN can be composed of (possibly many) convolutional layers, organized in blocks of layers, with each block typically operating on successively lower resolution versions of the image and control point data. Projection image 850 data and corresponding graphical aperture image 920 data can be introduced into the GAN by convolutional layers with a kernel of width one. Each convolution layer block can be composed of at least one convolution layer and a nonlinearity layer (e.g., a rectified linear unit or ReLU layer) that can apply a non-linear function to the output of the convolution layer. Each block may also contain a batch normalization layer, a scaling layer, or other layers that may be determined, such as to provide the most accurate estimates of desired graphical aperture image values. The last layer in the last block of each set can be a pooling layer that can downsample the convolution output data by an amount (e.g., one-half) and can take the maximum of the relevant layer-node values in the output convolution layer during a process known as max-pooling. The changing shapes of each convolution block set can indicate that a spatial resolution of the image data can be decreasing (block height) while the number of parameters per layer can be increasing (block width). The arrangement of layers and the layer-set compositions can preserve the information content of the DCNN.

During network training, the graphical aperture images output using the projection images 850 can be compared to the correct graphical aperture images 920 (e.g., training graphical aperture images), and the differences (e.g., errors) are used to correct the network parameters. The correction process is known as backpropagation. The network parameters can be the layer node weight coefficients and bias terms summarized as the Θ in FIG. 10 associated with the layer nodes whose values can be reset during backpropagation. Network training produces a network model or function f(X:Θ) indicated in FIG. 10. The model depends on both the input training images X and the network parameters Θ. Once trained, the network model can produce the graphical aperture images 1010 output at an output layer for a patient not used in training the network model. The graphical aperture images 1010 are the graphical aperture image estimates for a new patient.

FIG. 11A illustrates a flowchart of a process 1100A of exemplary operations for training a first generative model adapted for outputting a synthetic collection of graphical aperture images from an input collection of projection images. The process 1100A is illustrated from the perspective of a radiotherapy system 100, which trains and utilizes a first generative model using a GAN or CycleGAN as discussed in the preceding examples. However, corresponding operations may be performed by other devices or systems (including in offline training or verification settings separate from a particular image processing workflow or medical treatment).

As shown, a first phase of the flowchart workflow begins with operations (1110, 1120) to establish the parameters of training and model operations. The process 1100A begins with operations to receive (e.g., obtain, extract, identify) training image data (operation 1110) and constraints or conditions for training (operation 1120). In an example, this training image data may comprise image data from a plurality of human subjects, relating to a particular condition, anatomical feature, or anatomical area—such as pairs of projection images and real graphical aperture images of a target region for a subject and for various gantry angles. Also in an example, the constraints may relate to an imaging device, a treatment device, a patient, or medical treatment considerations. In an example, these constraints may include adversarial losses and cycle-consistency based losses.

The second phase of the process 1100A continues with training operations, including adversarial training of generative and discriminative models in a generative adversarial network (operation 1130). In an example, the adversarial training includes training the first generative model to generate simulated graphical aperture images by processing an input collection of projection images (operation 1142). The first collection of simulated graphical aperture images is provided to a discriminative model to train the discriminative model to classify the generated simulated collection of graphical aperture images as simulated or real training data (operation 1144). Also in this adversarial training, the output of the first generative model is used for training the discriminative model, and the output of the discriminative model is used for training the first generative model. As described in connection with FIG. 11B and process 1100B, the first collection of simulated graphical aperture images are respectively passed to a second generative model. The second generative model generates cycle projection images from the first collection of synthetic graphical aperture images. The cycle projection images are used in loss terms for training the first generative model. The training of the second generative model is illustrated as operations 1146 and 1148 in FIG. 11B.

In various examples, the first generative model and the discriminative model comprise respective convolutional neural networks. In further examples, the generative adversarial network is a cycle generative adversarial network where multiple generative and adversarial models are employed and the output from one generative model is provided as an input to a second generative model.

The process 1100A continues with the output of the first generative model for use in generating a synthetic collection of graphical aperture images (operation 1150), as the first generative model is adapted to generate a collection of graphical aperture images based on an input collection of projection images of a subject. In some cases, the output of the second generative model is used to generate a synthetic collection of projection images (operations 1152 and 1162 in FIG. 11B).

The process 1100A continues with the utilization of the trained first generative model to generate a synthetic collection of graphical aperture images (operation 1160) based on an input collection of projection images of a subject.

The process 1100A concludes with a final phase to implement updates to the first generative model, including updating the generative model based on additional training data (operation 1170) and outputting the updated trained first generative model (operation 1180). In some cases, a final phase is to implement updates to the second generative model, including updating the second generative model based on additional training data (operation 1172, FIG. 11B). In various examples, the updating may be produced in connection with the receipt of additional training image data and constraints (e.g., in a manner similar to operations 1110, 1120), or the performance of additional adversarial training (e.g., in a manner similar to operations 1130, 1142, 1144). In further examples, the generative model may be specifically updated based on approval, changes, or use of the collection of graphical aperture images (e.g., resulting from modification, verification, or changes to the image data by a medical professional). The flowchart concludes with the use of the updated trained generative model (operation 1190), such as may be performed in uses of the updated generative model for subsequent radiotherapy treatments.

FIG. 11B illustrates a flowchart of a process 1100B of exemplary operations for training a second generative model adapted for outputting a synthetic collection of projection images from an input collection of graphical aperture images (e.g., real or training graphical aperture images). The process 1100B is illustrated from the perspective of a radiotherapy system 100, which trains and utilizes a second generative model using a GAN or CycleGAN as discussed in the preceding examples. However, corresponding operations may be performed by other devices or systems (including in offline training or verification settings separate from a particular image processing workflow or medical treatment).

As shown, a first phase of the flowchart workflow begins with operations (1110, 1120) to establish the parameters of training and model operations. The process 1100B begins with operations to receive (e.g., obtain, extract, identify) training image data (operation 1110) and constraints or conditions for training (operation 1120). In an example, this training image data may comprise image data from a plurality of human subjects, relating to a particular condition, anatomical feature, or anatomical area—such as pairs of projection images and real graphical aperture images of a target region for a subject and for various gantry angles. Also in an example, the constraints may relate to an imaging device, a treatment device, a patient, or medical treatment considerations. In an example, these constraints may include adversarial losses and cycle-consistency based losses.

The second phase of the process 1100B continues with training operations, including adversarial training of generative and discriminative models in a generative adversarial network (operation 1130). In an example, the adversarial training includes training the second generative model to generate simulated projection images by processing an input collection of graphical aperture images (operation 1146). The first collection of simulated collection projection images is provided to a discriminative model to train the discriminative model to classify the generated simulated collection of projection images as simulated or real training data (operation 1148). Also in this adversarial training, the output of the second generative model is used for training the discriminative model, and the output of the discriminative model is used for training the second generative model. As described in connection with FIG. 11B and process 1100B, the first collection of simulated projection images are respectively passed to the first generator model. The first generative model generates cycle graphical aperture images from the first collection of synthetic projection images. The cycle graphical aperture images are used in loss terms for training the second generative model.

The process 1100B continues with the output of the second generative model for use in generating a synthetic collection of projection images (operation 1152), as the second generative model is adapted to generate a collection of projection images based on an input collection of graphical aperture images of a subject.

The process 100B continues with the utilization of the trained second generative model to generate a synthetic collection of projection images (operation 1162) based on an input collection of graphical aperture images of a subject.

The process 1100B concludes with a final phase to implement updates to the second generative model, including updating the second generative model based on additional training data (operation 1172) and outputting the updated trained second generative model. In various examples, the updating may be produced in connection with the receipt of additional training image data and constraints (e.g., in a manner similar to operations 1110, 1120), or the performance of additional adversarial training (e.g., in a manner similar to operations 1130, 1146, 1148). In further examples, the second generative model may be specifically updated based on approval, changes, or use of the collection of projection images (e.g., resulting from modification, verification, or changes to the image data by a medical professional).

FIG. 12 is a flowchart illustrating example operations of the image processing device 112 in performing process 1200, according to example embodiments. The process 1200 may be embodied in computer-readable instructions for execution by one or more processors such that the operations of the process 1200 may be performed in part or in whole by the functional components of the image processing device 112; accordingly, the process 1200 is described below by way of example with reference thereto. However, in other embodiments, at least some of the operations of the process 1200 may be deployed on various other hardware configurations. The process 1200 is therefore not intended to be limited to the image processing device 112 and can be implemented in whole, or in part, by any other component. Some or all of the operations of process 1200 can be in parallel, out of order, or entirely omitted.

At operation 1230, image processing device 112 receives a first projection image that represents a view of an anatomy of a subject from a first gantry angle of the radiotherapy treatment machine.

At operation 1250, image processing device 112 generates, using a generative model, a first graphical aperture image representation of MLC leaf positions at the first gantry angle based on the first projection image. The generative model is trained based on a generative adversarial network to process the projection images representing different views of a patient anatomy as input and provide respective graphical aperture image representations of the MLC leaf positions at different gantry angles corresponding to the different views as an output.

At operation 1260, image processing device 112 generates one or more radiotherapy equipment parameters based on the generated first graphical aperture image representation.

FIG. 13 illustrates a block diagram of an embodiment ofa machine 1300 on which one or more of the methods as discussed herein can be implemented. In one or more embodiments, one or more items of the image processing device 112 can be implemented by the machine 1300. In alternative embodiments, the machine 1300 operates as a standalone device or may be connected (e.g., networked) to other machines. In one or more embodiments, the image processing device 112 can include one or more of the items of the machine 1300. In a networked deployment, the machine 1300 may operate in the capacity of a server or a client machine in server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine may be a personal computer (PC), a tablet PC, a set-top box (STB), a Personal Digital Assistant (PDA), a cellular telephone, a web appliance, a network router, switch or bridge, or any machine capable of executing instructions (sequential or otherwise) that specify actions to be taken by that machine. Further, while only a single machine is illustrated, the term “machine” shall also be taken to include any collection of machines that individually or jointly execute a set (or multiple sets) of instructions to perform any one or more of the methodologies discussed herein.

The example machine 1300 includes processing circuitry or processor 1302 (e.g., a CPU, a graphics processing unit (GPJ), an ASIC, circuitry, such as one or more transistors, resistors, capacitors, inductors, diodes, logic gates, multiplexers, buffers, modulators, demodulators, radios (e.g., transmit or receive radios or transceivers), sensors 1321 (e.g., a transducer that converts one form of energy (e.g., light, heat, electrical, mechanical, or other energy) to another form of energy), or the like, or a combination thereof), a main memory 1304 and a static memory 1306, which communicate with each other via a bus 1308. The machine 1300 (e.g., computer system) may further include a video display device 1310 (e.g., a liquid crystal display (LCD) or a cathode ray tube (CRT)). The machine 1300 also includes an alphanumeric input device 1312 (e.g., a keyboard), a user interface (UI) navigation device 1314 (e.g., a mouse), a disk drive or mass storage unit 1316, a signal generation device 1318 (e.g., a speaker), and a network interface device 1320.

The disk drive unit 1316 includes a machine-readable medium 1322 on which is stored one or more sets of instructions and data structures (e.g., software) 1324 embodying or utilized by any one or more of the methodologies or functions described herein. The instructions 1324 may also reside, completely or at least partially, within the main memory 1304 and/or within the processor 1302 during execution thereof by the machine 1300, the main memory 1304 and the processor 1302 also constituting machine-readable media.

The machine 1300 as illustrated includes an output controller 1328. The output controller 1328 manages data flow to/from the machine 1300. The output controller 1328 is sometimes called a device controller, with software that directly interacts with the output controller 1328 being called a device driver.

While the machine-readable medium 1322 is shown in an embodiment to be a single medium, the term “machine-readable medium” may include a single medium or multiple media (e.g., a centralized or distributed database, and/or associated caches and servers) that store the one or more instructions or data structures. The term “machine-readable medium” shall also be taken to include any tangible medium that is capable of storing, encoding or carrying instructions for execution by the machine and that cause the machine to perform any one or more of the methodologies of the present disclosure, or that is capable of storing, encoding or carrying data structures utilized by or associated with such instructions. The term “machine-readable medium” shall accordingly be taken to include, but not be limited to, solid-state memories, and optical and magnetic media. Specific examples of machine-readable media include non-volatile memory, including by way of example semiconductor memory devices, e.g., Erasable Programmable Read-Only Memory (EPROM), EEPROM, and flash memory devices: magnetic disks such as internal hard disks and removable disks; magneto-optical disks, and CD-ROM and DVD-ROM disks.

The instructions 1324 may further be transmitted or received over a communications network 1326 using a transmission medium. The instructions 1324 may be transmitted using the network interface device 1320 and any one of a number of well-known transfer protocols (e.g., HTTP). Examples of communication networks include a LAN, a WAN, the Internet, mobile telephone networks, Plain Old Telephone (POTS) networks, and wireless data networks (e.g., WiFi and WiMax networks). The term “transmission medium” shall be taken to include any intangible medium that is capable of storing, encoding or carrying instructions for execution by the machine, and includes digital or analog communications signals or other intangible media to facilitate communication of such software.

As used herein, “communicatively coupled between” means that the entities on either of the coupling must communicate through an item therebetween and that those entities cannot communicate with each other without communicating through the item.

ADDITIONAL NOTES

The above detailed description includes references to the accompanying drawings, which form a part of the detailed description. The drawings show, by way of illustration but not by way of limitation, specific embodiments in which the disclosure can be practiced. These embodiments are also referred to herein as “examples.” Such examples can include elements in addition to those shown or described. However, the present inventors also contemplate examples in which only those elements shown or described are provided. Moreover, the present inventors also contemplate examples using any combination or permutation of those elements shown or described (or one or more aspects thereof), either with respect to a particular example (or one or more aspects thereof), or with respect to other examples (or one or more aspects thereof) shown or described herein.

All publications, patents, and patent documents referred to in this document are incorporated by reference herein in their entirety, as though individually incorporated by reference. In the event of inconsistent usages between this document and those documents so incorporated by reference, the usage in the incorporated reference(s) should be considered supplementary to that of this document; for irreconcilable inconsistencies, the usage in this document controls.

In this document, the terms “a,” “an,” “the,” and “said” are used when introducing elements of aspects of the disclosure or in the embodiments thereof, as is common in patent documents, to include one or more than one or more of the elements, independent of any other instances or usages of “at least one” or “one or more.” In this document, the term “or” is used to refer to a nonexclusive or, such that “A or B” includes “A but not B,” “B but not A,” and “A and B,” unless otherwise indicated.

In the appended claims, the terms “including” and “in which” are used as the plain-English equivalents of the respective terms “comprising” and “wherein.” Also, in the following claims, the terms “comprising,” “including,” and “having” are intended to be open-ended to mean that there may be additional elements other than the listed elements, such that after such a term (e.g., comprising, including, having) in a claim are still deemed to fall within the scope of that claim. Moreover, in the following claims, the terms “first,” “second,” and “third,” and so forth, are used merely as labels, and are not intended to impose numerical requirements on their objects.

Embodiments of the disclosure may be implemented with computer-executable instructions. The computer-executable instructions (e.g., software code) may be organized into one or more computer-executable components or modules. Aspects of the disclosure may be implemented with any number and organization of such components or modules. For example, aspects of the disclosure are not limited to the specific computer-executable instructions or the specific components or modules illustrated in the figures and described herein. Other embodiments of the disclosure may include different computer-executable instructions or components having more or less functionality than illustrated and described herein.

Method examples (e.g., operations and functions) described herein can be machine or computer-implemented at least in part (e.g., implemented as software code or instructions). Some examples can include a computer-readable medium or machine-readable medium encoded with instructions operable to configure an electronic device to perform methods as described in the above examples. An implementation of such methods can include software code, such as microcode, assembly language code, a higher-level language code, or the like (e.g., “source code”). Such software code can include computer-readable instructions for performing various methods (e.g., “object” or “executable code”). The software code may form portions of computer program products. Software implementations of the embodiments described herein may be provided via an article of manufacture with the code or instructions stored thereon, or via a method of operating a communication interface to send data via a communication interface (e.g., wirelessly, over the internet, via satellite communications, and the like).

Further, the software code may be tangibly stored on one or more volatile or non-volatile computer-readable storage media during execution or at other times. These computer-readable storage media may include any mechanism that stores information in a form accessible by a machine (e.g., computing device, electronic system, and the like), such as, but are not limited to, floppy disks, hard disks, removable magnetic disks, any form of magnetic disk storage media, CD-ROMS, magnetic-optical disks, removable optical disks (e.g., compact disks and digital video disks), flash memory devices, magnetic cassettes, memory cards or sticks (e.g., secure digital cards), RAMs (e.g., CMOS RAM and the like), recordable/non-recordable media (e.g., read only memories (ROMs)), EPROMS, EEPROMS, or any type of media suitable for storing electronic instructions, and the like. Such computer-readable storage medium is coupled to a computer system bus to be accessible by the processor and other parts of the OIS.

In an embodiment, the computer-readable storage medium may have encoded a data structure for treatment planning, wherein the treatment plan may be adaptive. The data structure for the computer-readable storage medium may be at least one of a Digital Imaging and Communications in Medicine (DICOM) format, an extended DICOM format, a XML format, and the like. DICOM is an international communications standard that defines the format used to transfer medical image-related data between various types of medical equipment. DICOM RT refers to the communication standards that are specific to radiation therapy.

In various embodiments of the disclosure, the method of creating a component or module can be implemented in software, hardware, or a combination thereof. The methods provided by various embodiments of the present disclosure, for example, can be implemented in software by using standard programming languages such as, for example, C, C++, Java, Python, and the like; and combinations thereof. As used herein, the terms “software” and “firmware” are interchangeable, and include any computer program stored in memory for execution by a computer.

A communication interface includes any mechanism that interfaces to any of a hardwired, wireless, optical, and the like, medium to communicate to another device, such as a memory bus interface, a processor bus interface, an Internet connection, a disk controller, and the like. The communication interface can be configured by providing configuration parameters and/or sending signals to prepare the communication interface to provide a data signal describing the software content. The communication interface can be accessed via one or more commands or signals sent to the communication interface.

The present disclosure also relates to a system for performing the operations herein. This system may be specially constructed for the required purposes, or it may comprise a general purpose computer selectively activated or reconfigured by a computer program stored in the computer. The order of execution or performance of the operations in embodiments of the disclosure illustrated and described herein is not essential, unless otherwise specified. That is, the operations may be performed in any order, unless otherwise specified, and embodiments of the disclosure may include additional or fewer operations than those disclosed herein. For example, it is contemplated that executing or performing a particular operation before, contemporaneously with, or after another operation is within the scope of aspects of the disclosure.

In view of the above, it will be seen that the several objects of the disclosure are achieved and other advantageous results attained. Having described aspects of the disclosure in detail, it will be apparent that modifications and variations are possible without departing from the scope of aspects of the disclosure as defined in the appended claims. As various changes could be made in the above constructions, products, and methods without departing from the scope of aspects of the disclosure, it is intended that all matter contained in the above description and shown in the accompanying drawings shall be interpreted as illustrative and not in a limiting sense.

The above description is intended to be illustrative, and not restrictive. For example, the above-described examples (or one or more aspects thereof) may be used in combination with each other. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the disclosure without departing from its scope. While the dimensions, types of materials and coatings described herein are intended to define the parameters of the disclosure, they are by no means limiting and are exemplary embodiments. Many other embodiments will be apparent to those of skill in the art upon reviewing the above description. The scope of the disclosure should, therefore, be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.

Also, in the above Detailed Description, various features may be grouped together to streamline the disclosure. This should not be interpreted as intending that an unclaimed disclosed feature is essential to any claim. Rather, inventive subject matter may lie in less than all features of a particular disclosed embodiment. Thus, the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separate embodiment. The scope of the disclosure should be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled. Further, the limitations of the following claims are not written in means-plus-function format and are not intended to be interpreted based on 35 U.S.C. § 112, sixth paragraph, unless and until such claim limitations expressly use the phrase “means for” followed by a statement of function void of further structure.

The Abstract is provided to comply with 37 C.F.R. § 1.72(b), to allow the reader to quickly ascertain the nature of the technical disclosure. It is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims.

Claims

1. A computer-implemented method comprising: receiving a set of pairs of image data for a plurality of gantry angles of a radiotherapy treatment machine, wherein a given pair of the set of pairs comprises a given projection image that represents a view of an anatomy of a subject from a given gantry angle and a given graphical aperture image of multi-leaf collimator (MLC) leaf positions at the given gantry angle; andtraining a neural network based on the set of pairs of image data for the plurality of gantry angles, wherein the trained neural network is used to predict an aperture image of MLC leaf positions for a specified gantry angle based on a new projection image.

2. The method of claim 1, wherein the neural network comprises a generative adversarial network (GAN), wherein the GAN comprises a conditional adversarial network (cGAN).

3. The method of claim 1, wherein the neural network comprises a cycle-consistent generative adversarial network (CycleGAN).

4. The method of claim 1, wherein the given gantry angle of a first pair of the set of pairs differs from the given gantry angle of a second pair of the set of pairs by a predetermined amount.

5. The method of claim 1, wherein: the neural network is configured to train a generative model using a discriminative model;values applied by the generative model and the discriminative model are established using adversarial training between the discriminative model and the generative model; andthe generative model and the discriminative model comprise respective convolutional neural networks.

6. The method of claim 5, wherein: the adversarial training comprises:training the generative model to generate a first synthetic graphical aperture image representation of the MLC leaf positions at a first gantry angle from a projection image that represents a view of a training subject anatomy from the first gantry angle; andtraining the discriminative model to classify the first synthetic graphical aperture image as a synthetic or a real training example graphical aperture image; andan output of the generative model is used for training the discriminative model and an output of the discriminative model is used for training the generative model.

7. The method of claim 6, wherein the neural network is trained using a cycle-consistent generative adversarial network (CycleGAN) comprising the generative model and the discriminative model, wherein the generative model is a first generative model and the discriminative model is a first discriminative model, wherein the CycleGAN further comprises: a second generative model trained to:process, from a given pair of the set of pairs, a given graphical aperture image representation of the MLC leaf positions at a given gantry angle as an input; andprovide a synthetic projection image that represents a view of a training subject anatomy from the given gantry angle as an output; anda second discriminative model trained to classify the synthetic projection image as a synthetic or a real projection image.

8. The method of claim 7, wherein the CycleGAN comprises a first portion to train the first generative model, the first portion being trained to: obtain a set of training projection images representing different views of a patient anatomy from prior treatments that are paired with training graphical aperture images corresponding to each of the different views, each of the training graphical aperture images being aligned with a respective one of the training projection images;transmit the set of training projection images to an input of the first generative model to output a first set of graphical aperture images;receive the first set of graphical aperture images at an input of the first discriminative model to classify the first set of graphical aperture images as a synthetic or real training set of graphical aperture images; andreceive the first set of graphical aperture images at an input of the second generative model to generate a first set of cycle projection images, for calculating cycle-consistency losses.

9. The method of claim 8, wherein the CycleGAN comprises a second portion that is trained to: transmit the training graphical aperture images corresponding to each of the different views to the input of the second generative model to output a first set of synthetic projection images;receive the first set of synthetic projection images at an input of the second discriminative model to classify the first set of synthetic projection images as synthetic or real training projection images; andreceive the first set of synthetic projection images at the input of the first generative model to generate a first set of cycle graphical aperture images for calculating cycle-consistency losses.

10. The method of claim 9, wherein: the cycle-consistency losses are generated based on a comparison of the first set of cycle projection images with the set of training projection images and a comparison of the first set of cycle graphical aperture images with the training graphical aperture images;the first generative model is trained to minimize or reduce a first loss term that represents an expectation of difference between a plurality of synthetic graphical aperture images and respectively paired training graphical aperture images; andthe second generative model is trained to minimize or reduce a second loss term that represents an expectation of difference between a plurality of synthetic projection images and respectively paired training projection images.

11. The method of claim 1, further comprising: obtaining radiotherapy treatment machine parameter information representing MLC leaf positions at gantry angles corresponding to a plurality of the views represented by the projection images of the set of pairs of image data and radiotherapy beam intensities corresponding to each of the projection images;generating training graphical aperture image representations based on the obtained radiotherapy treatment machine parameter information; andaligning each of the generated training graphical aperture image representations with the corresponding projection images.

12. The method of claim 11, wherein the training graphical aperture image representations and corresponding projection images are two-dimensional images or three-dimensional images comprising stacks of two-dimensional projection image and graphical aperture image pairs corresponding to an entire treatment fraction.

13. A system comprising: one or more processors configured to perform operations comprising: receiving a set of pairs of image data for a plurality of gantry angles of a radiotherapy treatment machine, wherein a given pair of the set of pairs comprises a given projection image that represents a view of an anatomy of a subject from a given gantry angle and a given graphical aperture image of multi-leaf collimator (MLC) leaf positions at the given gantry angle; andtraining a neural network based on the set of pairs of image data for the plurality of gantry angles, wherein the trained neural network is used to predict an aperture image of MLC leaf positions for a specified gantry angle based on a new projection image.

14. The system of claim 13, wherein the neural network comprises a generative adversarial network (GAN), wherein the GAN comprises a conditional adversarial network (cGAN) or a cycle-consistent generative adversarial network (CycleGAN).

15. The system of claim 13, wherein the given gantry angle of a first pair of the set of pairs differs from the given gantry angle of a second pair of the set of pairs by a predetermined amount.

16. The system of claim 13, wherein: the neural network is configured to train a generative model using a discriminative model;values applied by the generative model and the discriminative model are established using adversarial training between the discriminative model and the generative model; andthe generative model and the discriminative model comprise respective convolutional neural networks.

17. The system of claim 16, wherein: the adversarial training comprises:training the generative model to generate a first synthetic graphical aperture image representation of the MLC leaf positions at a first gantry angle from a projection image that represents a view of a training subject anatomy from the first gantry angle; andtraining the discriminative model to classify the first synthetic graphical aperture image as a synthetic or a real training example graphical aperture image; andan output of the generative model is used for training the discriminative model and an output of the discriminative model is used for training the generative model.

18. A non-transitory computer-readable medium comprising non-transitory computer-readable instructions for performing operations comprising: receiving a set of pairs of image data for a plurality of gantry angles of a radiotherapy treatment machine, wherein a given pair of the set of pairs comprises a given projection image that represents a view of an anatomy of a subject from a given gantry angle and a given graphical aperture image of multi-leaf collimator (MLC) leaf positions at the given gantry angle; andtraining a neural network based on the set of pairs of image data for the plurality of gantry angles, wherein the trained neural network is used to predict an aperture image of MLC leaf positions for a specified gantry angle based on a new projection image.

19. The non-transitory computer-readable medium of claim 18, wherein the neural network comprises a generative adversarial network (GAN), wherein the GAN comprises a conditional adversarial network (cGAN).

20. The non-transitory computer-readable medium of claim 18, wherein the neural network comprises a cycle-consistent generative adversarial network (CycleGAN).

21. The non-transitory computer-readable medium of claim 18, wherein the given gantry angle of a first pair of the set of pairs differs from the given gantry angle of a second pair of the set of pairs by a predetermined amount.

22. The non-transitory computer-readable medium of claim 18, wherein: the neural network is configured to train a generative model using a discriminative model;values applied by the generative model and the discriminative model are established using adversarial training between the discriminative model and the generative model; andthe generative model and the discriminative model comprise respective convolutional neural networks.

23. The non-transitory computer-readable medium of claim 22, wherein: the adversarial training comprises:training the generative model to generate a first synthetic graphical aperture image representation of the MLC leaf positions at a first gantry angle from a projection image that represents a view of a training subject anatomy from the first gantry angle; andtraining the discriminative model to classify the first synthetic graphical aperture image as a synthetic or a real training example graphical aperture image; andan output of the generative model is used for training the discriminative model and an output of the discriminative model is used for training the generative model.

24. A computer-implemented method comprising: receiving a projection image that represents a view of an anatomical region of interest;applying a neural network to the projection image to estimate an aperture image of multi-leaf collimator (MLC) leaf positions for a specified gantry angle corresponding to the view of an anatomical region of interest, the neural network being trained for a plurality of gantry angles of a radiotherapy treatment machine based on training data comprising image data for the plurality of gantry angles, wherein the image data comprises a given projection image that represents a view of an anatomy of a subject from a given gantry angle and a corresponding given graphical aperture image of MLC leaf positions at the given gantry angle; andgenerating a radiotherapy treatment plan based on the estimated aperture image of the MLC leaf positions.

25. The computer-implemented method of claim 24, further comprising generating one or more radiotherapy treatment machine parameters from the estimated aperture image.

26. The computer-implemented method of claim 25, further comprising computing a dose volume histogram or a three-dimensional dose distribution based on the generated one or more radiotherapy treatment machine parameters.

27. The computer-implemented method of claim 25, wherein the one or more radiotherapy treatment machine parameters include at least one of a gantry angle, an MLC jaw position, an MLC leaf position, or a radiation therapy beam intensity.

28. The computer-implemented method of claim 25, further comprising computing dosimetric parameters based on the generated one or more radiotherapy treatment machine parameters.

29. The computer-implemented method of claim 24, wherein the given projection image in the image data is generated by ray tracing or Fourier reconstruction.

30. A system comprising: one or more processors configured to perform operations comprising: receiving a projection image that represents a view of an anatomical region of interest;applying a neural network to the projection image to estimate an aperture image of multi-leaf collimator (MLC) leaf positions for a specified gantry angle corresponding to the view of an anatomical region of interest, the neural network being trained for a plurality of gantry angles of a radiotherapy treatment machine based on training data comprising image data for the plurality of gantry angles, wherein the image data comprises a given projection image that represents a view of an anatomy of a subject from a given gantry angle and a corresponding given graphical aperture image of MLC leaf positions at the given gantry angle; andgenerating a radiotherapy treatment plan based on the estimated aperture image of the MLC leaf positions.

31. The system of claim 30, further comprising operations for generating one or more radiotherapy treatment machine parameters from the estimated aperture image.

32. The system of claim 31, further comprising operations for computing a dose volume histogram or a three-dimensional dose distribution based on the generated one or more radiotherapy treatment machine parameters.

33. The system of claim 31, wherein the one or more radiotherapy treatment machine parameters include at least one of a gantry angle, an MLC jaw position, an MLC leaf position, or a radiation therapy beam intensity.

34. The system of claim 31, further comprising computing dosimetric parameters based on the generated one or more radiotherapy treatment machine parameters.

35. The system of claim 30, wherein the given projection image in the image data is generated by ray tracing or Fourier reconstruction.