Patent Application
20230360210
The present application claims priority under 35 U.S.C. § 119 to European Patent Application No. 22171617.8, filed May 4, 2022, the entire contents of which are incorporated herein by reference.
One or more example embodiments relates to a computer implemented method for generating a subject-specific map of a tissue property, in particular the magnetic susceptibility, of a region of interest within a subject, to various uses of such a magnetic susceptibility map, and a computer program and system.
In magnetic resonance imaging and spectroscopy, the homogeneity of the B0 magnetic field is of crucial importance. However, as soon as an object, such as a part of a subject's body, is introduced into the main magnetic field, the field lines are distorted. This distortion, also referred to as demagnetization field, is determined by the magnetic susceptibility distribution within the object. Especially the interface between air and tissue, as is found for example in the human head because of the sinuses, leads to particularly high magnetic field inhomogeneities. Although every magnetic resonance (MR) scanner has an active shimming system using coils with adjustable currents, usually the demagnetization field can be compensated only partially by shimming.
If the magnetic susceptibility distribution Chi(r) in the body part to be examined is known, as well as its positional orientation within the static magnetic field B0, then the effect of the demagnetization field can be corrected for, as shown for example in reference 9 (all references are listed at the end of the description).
Also, shimming generally does not allow for any dynamic changes of the orientation of the susceptibility distribution Chi(r) because of any subject motion. These effects may not be negligible and may lead to significant image degradation. In particular, for MR imaging sequences sensitive to off resonances, for example, gradient-echo echo planar imaging (GRE EPI) (see reference 7) or multi-TE GRE acquisitions as used for susceptibility weighted imaging (see reference 8), the effect of the demagnetization field needs to be taken into account.
However, at present no fast and sufficiently detailed methods for obtaining a subject-specific magnetic susceptibility distribution map are known.
One or more example embodiments provides a method for generating a subject-specific map of a tissue property, in particular magnetic susceptibility, of a region of interest within the subject. One or more example embodiments provides uses for such a magnetic susceptibility map.
The invention will now be illustrated by means of embodiments with reference to the attached drawings. In the drawings
According to one or more example embodiments, a computer implemented method for generating a subject-specific map of a tissue property, preferably the magnetic susceptibility, of a region of interest within the subject, the method comprising the steps of:
Apart from magnetic susceptibility, the tissue property may be an MR-related property, such as the MR relaxation times T1 or T2, or another electromagnetic property of the tissue, for example the dielectric permittivity, the electrical conductivity or the magnetic permeability.
One or more example embodiments of the present invention is based on the insight that the different tissue types in the human body have different susceptibility values, which however can be estimated in advance, as they are not subject-specific, or only to a small degree. Therefore, if an image is segmented into the various tissue types and typical susceptibility values for each tissue type are assigned to the segmented regions, a susceptibility map can be obtained. However, it has so far not been possible to automatically segment MR images in this way, because small and thin structures (such as the bones constituting the sinuses) as well as air and bone do not generate a large or any MR signal and therefore appear exactly the same in MR images, namely having very low pixel values within the noise range, which are generally depicted as black. Therefore, if an MR image is segmented into different tissue types, it is difficult to differentiate between small structures, bone and air, especially in the head. On the other hand, also attenuation correction maps (AC-maps) consist of only a limited number of tissue types, namely lungs, soft tissue, fat and bones. AC-maps are necessary for a correct reconstruction of PET volumes. Several methods have been proposed to estimate AC-maps from MR images, e.g., as described in reference 16, where an ultrashort echo time (UTE) sequence is used to derive the AC-map. Such MR images can be acquired simultaneously to the PET experiments with a PET-MR, instead of AC-maps derived from a separately acquired CT volume. However, these techniques, which provide a distribution of air cavities, bone and tissue, can still wrongly assign spatial positions to tissue instead of bone, or to tissue instead of air cavities, especially in the head.
One or more example embodiments of the present invention therefore uses an MR input image of the region of interest and applies an algorithm thereto, wherein an output image having improved contrast between bone and air is generated. According to one embodiment, the algorithm may comprise at least one trained artificial neural network (NN). The neural network may be a convolutional neural network, for example having 4 to 16 convolutional layers. The convolutional neural network may be based on an encoder-decoder structure with symmetric concatenations between corresponding states, for example of the type as described in reference 19. For example, each layer in the neural network may comprise one or more 3×3×3 kernels, batch normalisation, an activation function, for example a rectified linear unit, and a dropout layer in the encoding part, and vice versa in the decoding part. The NN may have been trained by a machine learning algorithm, for example by a back-propagation method. The NN may have been trained using real images, for example if the output image is a pseudo CT image, the NN may have been trained using MR images of a body part as input training data, and real CT images of the same body part as output training data.
According to a second embodiment, the algorithm generating an output image having improved contrast between bone and air may be an algorithm which is not based on a trained function, but which uses for example an anatomical atlas of the region of interest. An anatomical atlas is a standardised image showing a “typical” image or segmented image of the body part in question. Such an algorithm may for example comprise the step of segmenting the MR input image into air and one or several different tissues, and adapting (i.e. distorting) the anatomical atlas to fit the segmented image. Thereby, it is also possible to generate an output image having better contrast between bone and air.
The MR input image is preferably a three-dimensional (3D) image, although a stack of two-dimensional (2D) images may also be used to estimate the magnetic susceptibility map. The MR input image may also comprise several images of the region of interest, in particular several images which have been acquired in one imaging sequence, for example in-phase and opposed-phase images obtained by chemical shift imaging. The method of one or more example embodiments of the present invention may also be applied to a time series of MR input images, for example a series of EPI images which are acquired during a fMRI (functional magnetic resonance imaging) experiment, in which a series of images, such as EPI images, is acquired in order to observe changes in the subject's brain activity over time, in particular the change in BOLD-contrast (Blood Oxygen Level Dependent). In this case, the method of one or more example embodiments of the present invention may be used to obtain a time series of susceptibility maps and thus enable correction of the shimming or imaging parameters during the fMRI experiment, to ensure the best image quality.
In an embodiment, the neural network is trained on the specific body part examined, for example the human head, prostrate, kidney, abdomen, lung, heart of specific organs or limbs. The algorithm comprising the at least one neural network may therefore be chosen according to the body part in which the region of interest is situated. By this smart use of artificial intelligence, one or more example embodiments of the present invention can make use of (preferably convolutional) neural networks to improve the contrast between bone and air, which otherwise does not allow suitable segmentation of the images.
The output image is then segmented into air and at least two types of tissue, namely at least bone and at least one type of soft tissue (or soft tissue in general), to obtain a segmented image. In the simplest form, the segmentation differentiates between bone, air and soft tissue, wherein soft tissue comprises all types of soft tissue, e.g. fat, muscle, white matter, grey matter, skin, liver, etc. In more sophisticated embodiments, the soft tissue may also be differentiated into several different types, for example fat and other tissue. The segmentation step may use thresholding, 3D adaptive thresholding, model-based segmentation, atlas-based segmentation, histogram-based methods, edge detection, region-growing, or any other segmentation methods. The central assumption of model-based approaches is that the structures of interest have a tendency towards a particular shape. Therefore, one can seek a probabilistic model that characterizes the shape and its variation. When segmenting an image, constraints can be imposed using this model as a prior, as described in reference 38.
In the next step, predetermined values for the tissue property, preferably the magnetic susceptibility, are assigned to each pixel or voxel in the segmented image, according to its type of material (bone, soft tissue or air), to obtain the subject-specific map. The values used for the magnetic susceptibility can be for example −9.2 ppm for soft tissue, −8.44 ppm for bone and 0.36 ppm for air, as disclosed in reference 11 or 12. However, other predetermined values may also be used.
It has been shown that even this simplified map yields very good results in calculating the corresponding demagnetization field, since it is entirely subject-specific, without making use of any atlases or other preconceived ideas about the patient's anatomy. The only predetermined values are the susceptibility values of the different tissue types, which however are known and do not vary significantly from patient to patient.
According to one or more example embodiments, a computer implemented method for generating a subject-specific map of a tissue property, preferably the magnetic susceptibility, of a region of interest within the subject, the method comprising the steps of:
The algorithm may be a trained artificial neural network having the same properties and being trained in a similar manner as described herein.
One or more example embodiments of the present invention thereby provides an extremely fast and easily implemented method for obtaining a subject-specific map of magnetic susceptibility. Thereby, it may not be necessary in the future to measure field maps of the B0 field or the demagnetization field directly, as has been done in the past.
The method of one or more example embodiments of the present invention is preferably performed by a computer, which may be any processing unit such as a CPU and/or a GPU of a server or PC, or preferably a processing unit within an MR scanner. It may be performed on any computing device including a cloud computer, a laptop, tablet computer or other mobile device. The method is preferably carried out during an MR examination of a subject, i.e., while the subject is in the sensitive region of the MR scanner. This is because the generated tissue property map may be used to improve the quality of further MR acquisitions. The MR scanner may also be a PET/MR scanner or other hybrid device. The subject may be a human or animal, in particular a patient during an MRI examination.
According to an embodiment, the output image of the neural network is a synthetic attenuation correction map. The advantage of using attenuation correction maps (AC-maps) is that they generally consist of only a limited number of tissue types, namely bone, fat and soft tissue. Therefore, they may be particularly useful in generating magnetic susceptibility maps, which also require only a limited number of tissue types to provide sufficient correction for field distortions and motion. Also, attenuation correction maps can be estimated from MR images. In recent years, trained convolution neural networks were proposed to obtain better AC-maps than atlas-based methods, e.g., from high resolution and fast Dixon acquisitions, which can be measured within 39 seconds, or from UTE volumes with two echo times, as described in references 19 and 29. The resulting AC-maps have been demonstrated to have a more accurate distribution of bones and air cavities in the head than previous atlas-based methods. The obtained AC-maps also have a high resolution, high enough to show the detailed anatomical structure in the sinuses and to differentiate between bone and air. With trained neural networks, it is possible to deliver high-resolution attenuation correction maps from MR images.
According to another embodiment, the output image of the neural network is a synthetic CT (Computed Tomography) image. Such a method for transforming a low-resolution MR input image into a high-resolution MR image using a CNN based network, and transforming the high-resolution MR image to a pseudo-CT image has been proposed in US 2020/003 4948 A1 (reference 20). In this method, two convolutional neural networks are used one after the other, and the output image is again suitable for segmenting into the different tissue types, because CT images distinguish very well between bone and air, so a segmentation into soft tissue, bone and air does not provide any problems. Also, CT images have a high resolution, and a CNN may be trained to achieve such high resolution, even from lower resolution MR input images.
The MR input image of the region of interest may be acquired by any possible MR sequence, for example spin echo, fast spin echo, gradient echo, echo planar imaging (EPI) etc. The MR input image may be three-dimensional (3D) or two-dimensional (2D), wherein the method may also be applied to a stack of 2D images. The MR input image may also comprise several images, such as in-phase and opposed-phase, or a time series of images. According to an embodiment, the MR input image has been acquired using a fast or a standard MR imaging protocol, preferably Dixon, MPRAGE, MP2RAGE, MP3RAGE or sequences with ultrashort echo time (UTE). The MR input image can be an image that is acquired also for diagnostic purposes during a patient examination. It is advantageous if the MR input image is acquired within less than 3 minutes, preferably 30 seconds to 2 minutes, and the above-mentioned imaging methods allow an acquisition of a 3D image volume in about 10-60 seconds, preferably 20-40 seconds. A high-resolution 3D Dixon VIBE protocol may be used with CAIPIRINHA acceleration. As an example, such a protocol may take 39 seconds. The acquisition may be made even shorter if an acquisition with Compressed Sensing is used, and/or the resolution is reduced. A further advantage of UTE, MPRAGE and Dixon sequences is that they are available for MR scanners at all field strengths ranging from 1.5 T to 7 T, for example for 1.5 T, 3 T and 7 T MR scanners. A further advantage is that MR input images acquired using UTE and Dixon at 3 T have a good suitability to be converted for example into AC-maps using convolutional neural networks, as demonstrated in references 19 and 29.
According to an embodiment, the method comprises the further steps of receiving a morphological image of the region of interest, segmenting the morphological image into at least two categories of soft tissue types to obtain a soft tissue segmentation image, and combining the soft tissue segmentation image with the segmented image to obtain an improved segmented image.
A morphological image may be a morphological MR image, in particular it may be the (same) MR input image to which the algorithm, e.g. the trained artificial neural network, is applied to generate the image with improved contrast between bone and air. It may also be another MR image, for example one with better contrast between different soft tissues. The morphological image may be acquired in the same examination session than the MR input image, but may also have been acquired previously and may then be registered with the MR input image. A morphological MR image may for example have been acquired by MPRAGE (Magnetization Prepared RApid Gradient Echo), which uses a non-selective inversion pulse followed by a collection of rapidly acquired gradient echoes obtained at short echo times and small flip angles, or by MP2RAGE or MP3RAGE. The morphological image may also be acquired using spin echo, fast spin echo, fast gradient echo (e.g. bSSFP), FLASH or other MR sequences allowing a good contrast for example between grey matter and white matter in the human brain.
In the next step, the morphological image is segmented into at least two types of soft tissue, for example, grey matter, white matter and cerebral spinal fluid (CSF) in the brain. In addition, in other body areas, the types of soft tissue may include one or several of fat, muscle, liver tissue, cartilage, blood, skin or lung. In MR imaging of the brain, typically the soft tissue will be segmented into at least grey matter, white matter and CSF. Thereby, the magnetic susceptibility can be estimated more accurately for soft tissue. The combination may be performed by replacing the soft tissue segmentation in the segmented image with the more detailed soft tissue segmentation from the soft tissue segmentation image. In the next step, predetermined values for the tissue property are assigned to each type of tissue and soft tissue in the improved segmented image to obtain the subject-specific map of the tissue property, preferably the magnetic susceptibility.
The tissue segmentation of the morphological images may be performed using thresholding methods or other segmentation method as described above, but also using neural networks which are trained to label different soft tissue types.
In an alternative, the soft tissue segmentation image may also be assigned values for the magnetic susceptibility directly, and then the two magnetic susceptibility maps which are ultimately derived from the MR input image or from the MR input image and the morphological image may be combined. The combination may be done by applying a mask to the soft tissue and using the values derived from the morphological image for the soft tissue and for all other image areas the values derived from the output image of the algorithm.
According to an embodiment, the region of interest is at least partially within the head of the subject. As mentioned above, the human head poses particular challenges to MR protocols which are sensitive to be B0 field inhomogeneities, because of the many air cavities in the head. Therefore, the use of an accurate magnetic susceptibility map can yield significant improvements in image quality. In other embodiments, the region of interest may at least partially include the human prostate, the kidney, the heart, or indeed any other organ or body part.
According to an embodiment, the tissue property is the magnetic susceptibility and the method comprises the further step of estimating a map of the demagnetization field caused by the subject when placed in a B0 magnetic field, wherein the estimation is done based on the magnetic susceptibility map. The demagnetization field Bd is the magnetic field generated by an object with a given magnetisation and is related to the susceptibility distribution of the object and the static magnetic field strength. It can be estimated by the convolution of the magnetic susceptibility distribution/map with the magnetic dipole and the Lorentz sphere correction, as described in references 3 and 4. Similar methods may be used to calculate a map of the demagnetization field from the susceptibility map and the B0 magnetic field, as described for example in reference 12. This corresponds to a calculation of the magnetic field inhomogeneity induced by the presence of the part of the subject in the B0 field. The demagnetization field map depends also on the exact orientation of the susceptibility map within the field of view.
The proposed invention can make use of convolutional neural networks (CNN's) and the possibility to segment the tissue types of interest, as described in references 9 and 11, and assigning them their corresponding susceptibility values. This approach is then much simpler and quicker than to perform the method proposed in reference 11. It only requires moderate computational power in a reasonable amount of time, wherein a modern GPU is already sufficient. The proposed approach may avoid using an atlas, it only relies on MR measurements from one or several acquisitions.
In the following, novel uses of a thus obtained magnetic susceptibility map are described. The different uses described herein may be applied cumulatively in one MR imaging session.
Generally, according to one or more example embodiments, the use of a magnetic susceptibility map generated by the method described above improves the quality of further MR acquisition from the region of interest. Preferably, these further MR acquisitions take place in the same MR examination session, although the magnetic field map may also be used in later MR examination sessions, once it has been registered with the position of the respective body part during that later session.
According to an embodiment, the magnetic suitability map is used to improve the homogeneity of a B0 magnetic field of an MR scanner in which the subject is placed for MR acquisition, by estimating a map of the demagnetization field caused by the subject when placed in the B0 magnetic field from the magnetic susceptibility map, and adjusting the shim coefficients of the MR scanner to compensate for the demagnetization field. Thereby, the standard shimming using the shim coils is improved, because the demagnetising field map contains information also in regions where no MR signal is available and has potentially a higher resolution than the field maps which are usually acquired in the context of shimming. Thus, more voxels can be used for fitting the shim coefficients.
In a further embodiment, the quality of MR acquisition from the region of interest may be improved by tracking any motion of the region of interest during the acquisition of an MR input image, applying the tracked motion to the magnetic susceptibility map to obtain a dynamic magnetic susceptibility map, estimating a dynamic demagnetization field map from the dynamic magnetic susceptibility map, and adjusting the shim coefficients in real time to compensate for the dynamic demagnetization field.
Subject motion is a major problem for magnetic resonance imaging. Prospective motion correction methods use an external device to track the subject's position, for example a tracking target comprising reflective markers and which are being watched by tracking cameras, as described in reference 1. Such a tracking system is capable of reporting positions and orientations of several tracking targets attached to the patient and thereby record or track patient motion in real time. This information may be used to update the position of the imaging volume prior to every excitation of the spin system. Other methods for tracking the motion of the region of interest include image-based tracking, such as described in reference 2. According to this method, a rigid body estimation of head movements is obtained by image-based motion detection. Both of these methods allow to follow the position and orientation of the region of interest during the acquisition of an MR image. According to an embodiment, this motion, which may preferably be a rigid motion including three translational and three rotational degrees of freedom, but which may also have more degrees of freedom, may be applied to the magnetic susceptibility map during the MR acquisition to obtain a dynamic magnetic susceptibility map. In this context, “dynamic” means that the map is available for different time points, i.e. it is a time sequence of magnetic susceptibility maps which is generated for the time span of the MR input image acquisition, during which the motion is tracked. For example, the tracked translational and rotational movements are imposed on the susceptibility map at each time point during the MR acquisition, with a pre-determined temporal resolution. Preferably in real time, e.g. at the pre-determined temporal resolution, a dynamic demagnetization field map is estimated from the dynamic magnetic susceptibility map, by the methods described herein. This map of the demagnetization field therefore takes into account any patient motion and is preferably generated in real time during the MR input image acquisition. According to this aspect of one or more example embodiments of the present invention, it may be used to adjust the shim coefficients in real time to compensate for the respective dynamic demagnetization field. Thereby, an excellent image quality is obtained even for moving subjects.
According to another embodiment, the magnetic susceptibility map may be used for motion correction of an MR image acquired with an imaging sequence from the region of interest by tracking any motion of the region of interest during the acquisition of an MR image; applying the tracked motion to the magnetic susceptibility map to obtain a dynamic magnetic susceptibility map; estimating a dynamic demagnetization field map from the dynamic magnetic susceptibility map; and adjusting the imaging sequence in real time to compensate for the dynamic demagnetization field. Again, the motion tracking may be performed either using image-based tracking or using an external tracking device, for example an optical tracking device. The further steps including generating a dynamic magnetic susceptibility map and therefore a dynamic demagnetization field may be performed as described above. In the final step, the imaging sequence is adjusted in real time to compensate for the dynamic magnetisation field. In other words, the dynamic demagnetization field map can be used for dynamic geometric distortion correction, for example for EPI and other sequences sensitive to the inhomogeneous field and its changes, using rotations and translations for body regions which are moving during the MR exam, and which information may be made available by the herein described methods. The encountered motion during the acquisition can be applied to the susceptibility distribution. More complex nonrigid deformations for the body region of interest may also be tracked. Also in this case, the shim magnetic field may be taken into account.
According to a further embodiment, a dynamic demagnetization field map may be obtained as described herein above, and may be used for retrospective motion correction of an acquired MR image, which may be the MR input image, but may also be an MR image acquired afterwards during the same patient examination from the same or overlapping field of view. A suitable method is for example described in reference 22, WO 2011/127942 A1. This patent also uses a computed field inhomogeneity, which however is calculated from a susceptibility model which is less detailed than the susceptibility map proposed herein. The dynamic magnetisation field may be in particular used for correcting image distortions and/or intensity modulations. Thereby, the image quality of MR images of moving subjects may be improved. Using this method, even EPI imaging may be used, which is very sensitive to B0 distortions. If a time series of MR images is acquired, the dynamic demagnetization field map may be used to correct the geometric image distortions at each time point during the acquisition, in particular for each MR image in the time series.
According to another embodiment, the magnetic susceptibility map may be used to improve the quality of dynamic chemical exchange saturation transfer (CEST) image acquisition of the region of interest by tracking any motion of the region of interest, applying the tracked motion to the magnetic susceptibility map, estimating a dynamic demagnetization field map and correcting the dynamic CEST images by using the dynamic demagnetization field map. Since CEST imaging is based on the continuous transfer of excited 1H protons from one chemical species to the surrounding water, leading to a build-up of saturation in water, it requires frequency selective excitation of protons and therefore a high homogeneity of the B0 magnetic field. A knowledge of the demagnetization field map can be used to shift back the z-spectrum. In other words, knowledge of the off-resonance frequency shift per voxel is required so that all the measurements can be shifted back to their designated spectral positions.
According to an embodiment, the magnetic susceptibility map can also be used to improve the quality of dynamic CEST image acquisition of the region of interest by tracking any motion of the region of interest during the acquisition; applying the tracked motion to the magnetic susceptibility map to obtain a dynamic magnetic susceptibility map; estimating a dynamic demagnetization field from the dynamic magnetic susceptibility map; and correcting the dynamic Chemical Exchange Saturation Transfer images by using the dynamic demagnetization field map.
In other words, the inventive susceptibility map also allows to improve the quality of dynamic CEST images. Z-spectra estimated voxel-wise from CEST acquisitions, typically gradient echo acquisitions, need to be corrected for B0 inhomogeneity at the corresponding voxel. When dynamic CEST experiments are performed, as for example in glucose CEST described in references 24 and 25, then it is necessary to take the time variations of the magnetic inhomogeneity field into account, which can be dependent of subject motion. Again, an accurate subject-specific susceptibility model which can be varied with the subject's position will improve the correction for the field inhomogeneity.
One or more example embodiments of the present invention is further directed to the use of a magnetic susceptibility map in the calculation of radio-frequency (RF) excitation pulses of a multi-transmit system of an MR scanner for a given excitation trajectory for parallel excitation. In parallel excitation, a transmit system having several RF antennas (i.e. a multi-transmit system), for example an array or phased array of RF coils, is used to produce a complex RF pulse. For the same RF pulses, gradient fields are applied during the RF excitation, allowing specific desired trajectories of the magnetisation in transmit k-space. To realize these trajectories, an accurate knowledge of the magnetic field is important, see reference 31. Therefore, a subject-specific demagnetization field can be used to improve the calculation of RF excitation pulses.
According to a further embodiment, the subject-specific demagnetization field map can be used instead of a standard field map in advanced image reconstruction methods. Thus, after the method for generating a subject-specific map of a tissue property has been carried out, a map of the demagnetization field is calculated from the magnetic susceptibility map. Optionally, a further MR image is then acquired while the subject is placed with the MR Scanner, and the demagnetization field is used in the reconstruction of the acquired image. Preferably, image reconstruction involves advanced image reconstruction methods, such as iterative image reconstruction using an MR forward model, as described for example in reference 33 or 35. The image reconstruction method of ref. 35 includes a retrospective motion correction by including motion operations into the MR forward model, wherein the MR forward model is essentially a mathematical description (e.g. an encoding matrix) of the operation by which the MR image is generated from the proton distribution. In case of multi-channel k-space data acquired in a multi-shot imaging protocol, the MR forward model may be described by an encoding matrix, which includes the effects of rigid-body motion for each shot, Fourier encoding, and optionally subsampling and coil sensitivities of the multi-channel coil array. The solving of such an MR forward model usually requires several optimization iterations. The SENSE (SENSitivity Encoding) forward model was introduced by reference 36. According to an embodiment of the present invention, an MR forward model used in advanced, in particular iterative, image reconstruction may include the effect of the demagnetization field calculated from a susceptibility map, which has been generated according to the inventive method. This reduces the geometric distortions of the reconstructed image.
In another embodiment, the demagnetization field may be used in the reconstruction of MR data acquired using compressed sensing (e.g. described in reference 37) to improve the reconstructed images with regard to geometric distortion, as described in reference 5. Compressed sensing typically involves randomized subsampling of k-space.
According to a further embodiment, the subject-specific demagnetization field map can be used in quantitative susceptibility mapping, in particular in one or several steps of the post-processing pipeline for quantitative susceptibility mapping (QSM), which are summarized in references [26] and [27]. Typically (but not limited to), a 3D multi-TE GRE acquisition is used for the acquisition and no signal is available in regions consisting of bones. The QSM pipeline can be separated in several steps. One of the steps comprises removing the background field of the frequency map obtained from the acquired data over a masked region of the brain or the body part in question (the masked region is defined where MR signal is sufficient, e.g. above noise level) to obtain a background field corrected frequency map. The subject-specific demagnetization field map in this step can lead to an improved background field corrected frequency map, because magnetic field information outside of the brain or the body part is available. In a final step of the QSM post-processing pipeline, the susceptibility distribution is estimated from the masked background field corrected frequency map which can be achieved numerically by solving the ill-posed problem of deconvoluting the masked background field corrected frequency map with the unit magnetic dipole response. The additional knowledge of having a subject-specific susceptibility distribution helps to improve the condition of the ill-posed problem by, for example, generating a simulated demagnetization field in at least two additional different orientations of the susceptibility distribution to the acquired 3D multi-TE GRE image, similar as in reference [34]. Another possibility is to further constrain the inverse ill-posed problem by including the susceptibility distribution in the numerical algorithm.
One or more example embodiments of the present invention is also directed to a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method, either the method for generating a subject-specific map of a tissue property, or using such map in one of the described uses. The computer may be any processing unit, cloud computer or server, also a processing unit on a mobile device such as a laptop, smartphone or tablet.
One or more example embodiments of the present invention is also directed to a non-transient digital storage medium on which the computer program is stored. Such digital storage medium may be a hard disk, optical disk, magnetic storage medium, SD card, USB stick, CD-ROM or any possible portable or nonportable data storage medium.
One or more example embodiments of the present invention is also directed to a system for generating a subject-specific map of a tissue property, in particular the magnetic susceptibility, of a region of interest within a subject. The system comprises a first interface, configured for receiving an MR input image of the region of interest; a computational unit, configured for feeding the MR image as input into an algorithm, wherein the output of the algorithm is an output image having improved contrast between bone and air; segmenting the output image into air and at least two types of tissue, namely at least bone and at least one type of soft tissue, to obtain a segmented image; assigning pre-determined values for the tissue property to each category of tissue and air in the segmented image to obtain the subject-specific map of the tissue property; and a second interface, configured for providing the map of the tissue property.
The system may be part of an MR scanner, and particular part of the computer controlling the MR scanner and the image acquisition. However, the system may also be a remote computer to which the MR image is transferred.
Similar features are designated with the same reference numbers in the following.
The susceptibility map 20 may also be used for dynamic corrections, as illustrated in
It will be understood that, although the terms first, second, etc. may be used herein to describe various elements, components, regions, layers, and/or sections, these elements, components, regions, layers, and/or sections, should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. As used herein, the term “and/or,” includes any and all combinations of one or more of the associated listed items. The phrase “at least one of” has the same meaning as “and/or”.
Spatially relative terms, such as “beneath,” “below,” “lower,” “under,” “above,” “upper,” and the like, may be used herein for ease of description to describe one element or feature's relationship to another element(s) or feature(s) as illustrated in the figures. It will be understood that the spatially relative terms are intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if the device in the figures is turned over, elements described as “below,” “beneath,” or “under,” other elements or features would then be oriented “above” the other elements or features. Thus, the example terms “below” and “under” may encompass both an orientation of above and below. The device may be otherwise oriented (rotated 90 degrees or at other orientations) and the spatially relative descriptors used herein interpreted accordingly. In addition, when an element is referred to as being “between” two elements, the element may be the only element between the two elements, or one or more other intervening elements may be present.
Spatial and functional relationships between elements (for example, between modules) are described using various terms, including “on,” “connected,” “engaged,” “interfaced,” and “coupled.” Unless explicitly described as being “direct,” when a relationship between first and second elements is described in the disclosure, that relationship encompasses a direct relationship where no other intervening elements are present between the first and second elements, and also an indirect relationship where one or more intervening elements are present (either spatially or functionally) between the first and second elements. In contrast, when an element is referred to as being “directly” on, connected, engaged, interfaced, or coupled to another element, there are no intervening elements present. Other words used to describe the relationship between elements should be interpreted in a like fashion (e.g., “between,” versus “directly between,” “adjacent,” versus “directly adjacent,” etc.).
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the terms “and/or” and “at least one of” include any and all combinations of one or more of the associated listed items. It will be further understood that the terms “comprises,” “comprising,” “includes,” and/or “including,” when used herein, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. As used herein, the term “and/or” includes any and all combinations of one or more of the associated listed items. Expressions such as “at least one of,” when preceding a list of elements, modify the entire list of elements and do not modify the individual elements of the list. Also, the term “example” is intended to refer to an example or illustration.
It should also be noted that in some alternative implementations, the functions/acts noted may occur out of the order noted in the figures. For example, two figures shown in succession may in fact be executed substantially concurrently or may sometimes be executed in the reverse order, depending upon the functionality/acts involved.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which example embodiments belong. It will be further understood that terms, e.g., those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
It is noted that some example embodiments may be described with reference to acts and symbolic representations of operations (e.g., in the form of flow charts, flow diagrams, data flow diagrams, structure diagrams, block diagrams, etc.) that may be implemented in conjunction with units and/or devices discussed above. Although discussed in a particularly manner, a function or operation specified in a specific block may be performed differently from the flow specified in a flowchart, flow diagram, etc. For example, functions or operations illustrated as being performed serially in two consecutive blocks may actually be performed simultaneously, or in some cases be performed in reverse order. Although the flowcharts describe the operations as sequential processes, many of the operations may be performed in parallel, concurrently or simultaneously. In addition, the order of operations may be re-arranged. The processes may be terminated when their operations are completed, but may also have additional steps not included in the figure. The processes may correspond to methods, functions, procedures, subroutines, subprograms, etc.
Specific structural and functional details disclosed herein are merely representative for purposes of describing example embodiments. The present invention may, however, be embodied in many alternate forms and should not be construed as limited to only the embodiments set forth herein.
In addition, or alternative, to that discussed above, units and/or devices according to one or more example embodiments may be implemented using hardware, software, and/or a combination thereof. For example, hardware devices may be implemented using processing circuity such as, but not limited to, a processor, Central Processing Unit (CPU), a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a System-on-Chip (SoC), a programmable logic unit, a microprocessor, or any other device capable of responding to and executing instructions in a defined manner. Portions of the example embodiments and corresponding detailed description may be presented in terms of software, or algorithms and symbolic representations of operation on data bits within a computer memory. These descriptions and representations are the ones by which those of ordinary skill in the art effectively convey the substance of their work to others of ordinary skill in the art. An algorithm, as the term is used here, and as it is used generally, is conceived to be a self-consistent sequence of steps leading to a desired result.
It should be borne in mind that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. Unless specifically stated otherwise, or as is apparent from the discussion, terms such as “processing” or “computing” or “calculating” or “determining” of “displaying” or the like, refer to the action and processes of a computer system, or similar electronic computing device/hardware, that manipulates and transforms data represented as physical, electronic quantities within the computer system's registers and memories into other data similarly represented as physical quantities within the computer system memories or registers or other such information storage, transmission or display devices.
In this application, including the definitions below, the term ‘interface’ may be replaced with the term ‘circuit.’
The module may include one or more interface circuits. In some examples, the interface circuits may include wired or wireless interfaces that are connected to a local area network (LAN), the Internet, a wide area network (WAN), or combinations thereof. The functionality of any given module of the present disclosure may be distributed among multiple modules that are connected via interface circuits. For example, multiple modules may allow load balancing. In a further example, a server (also known as remote, or cloud) module may accomplish some functionality on behalf of a client module.
Software may include a computer program, program code, instructions, or some combination thereof, for independently or collectively instructing or configuring a hardware device to operate as desired. The computer program and/or program code may include program or computer-readable instructions, software components, software modules, data files, data structures, and/or the like, capable of being implemented by one or more hardware devices, such as one or more of the hardware devices mentioned above. Examples of program code include both machine code produced by a compiler and higher level program code that is executed using an interpreter.
For example, when a hardware device is a computer processing device (e.g., a processor, Central Processing Unit (CPU), a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a microprocessor, etc.), the computer processing device may be configured to carry out program code by performing arithmetical, logical, and input/output operations, according to the program code. Once the program code is loaded into a computer processing device, the computer processing device may be programmed to perform the program code, thereby transforming the computer processing device into a special purpose computer processing device. In a more specific example, when the program code is loaded into a processor, the processor becomes programmed to perform the program code and operations corresponding thereto, thereby transforming the processor into a special purpose processor.
Example embodiments may be described with reference to acts and symbolic representations of operations (e.g., in the form of flow charts, flow diagrams, data flow diagrams, structure diagrams, block diagrams, etc.) that may be implemented in conjunction with units and/or devices discussed in more detail below. Although discussed in a particularly manner, a function or operation specified in a specific block may be performed differently from the flow specified in a flowchart, flow diagram, etc. For example, functions or operations illustrated as being performed serially in two consecutive blocks may actually be performed simultaneously, or in some cases be performed in reverse order.
Units and/or devices according to one or more example embodiments may also include one or more storage devices. The one or more storage devices may be tangible or non-transitory computer-readable storage media, such as random access memory (RAM), read only memory (ROM), a permanent mass storage device (such as a disk drive), solid state (e.g., NAND flash) device, and/or any other like data storage mechanism capable of storing and recording data. The one or more storage devices may be configured to store computer programs, program code, instructions, or some combination thereof, for one or more operating systems and/or for implementing the example embodiments described herein. The computer programs, program code, instructions, or some combination thereof, may also be loaded from a separate computer readable storage medium into the one or more storage devices and/or one or more computer processing devices using a drive mechanism. Such separate computer readable storage medium may include a Universal Serial Bus (USB) flash drive, a memory stick, a Blu-ray/DVD/CD-ROM drive, a memory card, and/or other like computer readable storage media. The computer programs, program code, instructions, or some combination thereof, may be loaded into the one or more storage devices and/or the one or more computer processing devices from a remote data storage device via a network interface, rather than via a local computer readable storage medium. Additionally, the computer programs, program code, instructions, or some combination thereof, may be loaded into the one or more storage devices and/or the one or more processors from a remote computing system that is configured to transfer and/or distribute the computer programs, program code, instructions, or some combination thereof, over a network. The remote computing system may transfer and/or distribute the computer programs, program code, instructions, or some combination thereof, via a wired interface, an air interface, and/or any other like medium.
The one or more hardware devices, the one or more storage devices, and/or the computer programs, program code, instructions, or some combination thereof, may be specially designed and constructed for the purposes of the example embodiments, or they may be known devices that are altered and/or modified for the purposes of example embodiments.
The computer programs include processor-executable instructions that are stored on at least one non-transitory computer-readable medium (memory). The computer programs may also include or rely on stored data. The computer programs may encompass a basic input/output system (BIOS) that interacts with hardware of the special purpose computer, device drivers that interact with particular devices of the special purpose computer, one or more operating systems, user applications, background services, background applications, etc. As such, the one or more processors may be configured to execute the processor executable instructions.
The apparatuses and methods described in this application may be partially or fully implemented by a special purpose computer created by configuring a general purpose computer to execute one or more particular functions embodied in computer programs. The functional blocks and flowchart elements described above serve as software specifications, which can be translated into the computer programs by the routine work of a skilled technician or programmer.
Although described with reference to specific examples and drawings, modifications, additions and substitutions of example embodiments may be variously made according to the description by those of ordinary skill in the art. For example, the described techniques may be performed in an order different with that of the methods described, and/or components such as the described system, architecture, devices, circuit, and the like, may be connected or combined to be different from the above-described methods, or results may be appropriately achieved by other components or equivalents.
| Number | Date | Country | Kind |
|---|---|---|---|
| 22171617.8 | May 2022 | EP | regional |
