Patent Application
20130101197
1. Field of the Invention
This invention is directed to a method and apparatus for generating an enhanced image for display from medical imaging data of a subject, in particular for elongated features of interest in an image volume.
2. Description of the Prior Art
The definition of regions or volumes of interest (ROI/VOI) is a typical precursor to quantitative analysis of medical images, such as nuclear medicine emission images (for example, PET or SPECT). Such regions may be defined around areas of high intensity which correspond to high tracer uptake (hotspots). For example, in FDG-PET images for oncology, such areas may be indicative of the presence of a tumor. Oncology physicians frequently annotate lesions in PET scans for the purpose of making a diagnosis, or for use in radiotherapy. The mean or maximum tracer uptake can aid a reader in determining the likelihood of cancer. In longitudinal studies, considering the change in intensity or uptake on corresponding VOIs from images at different temporal stages may be used to determine whether a tumor has regressed.
In oncology the identification, diagnosis, follow-up, and reporting of lesions are key steps in the clinical routine. Bone metastases (in the ribs and spine, for example) occur in many cancer types. The PET tracer F-18 sodium fluoride (18F—NaF), which features an increased uptake in osteolytic and osteoblastic bone lesions, can be used for imaging bone lesions. It has been shown that twice as many benign as well as malignant lesions will be visualized on 18F—NaF PET bone imaging versus conventional 99 mTc phosphonate SPECT or planar imaging. As a consequence, 18F—NaF PET results changed clinical management in about 18% of patients.
The detection and reporting of bone lesions can be cumbersome, particularly for patients with multiple lesions or wide-spread metastases. As a consequence, the manual detection and reporting of bone lesions is often time consuming and hence also prone to errors. In case of multiple lesions, e.g., at different locations along different ribs, the individual locations are frequently not properly documented due to time restrictions in clinical routine.
PET/SPECT bone scans are typically read on either a slice-by-slice basis (axial, coronal, sagittal) or by means of 3D MIPs. Each finding then needs to be manually reported (left/right rib, rib number, position along rib). However, after identification of a lesion along a particular rib, this information needs to be manually obtained by, e.g., counting the ribs visible in the 3D volume. As a consequence, detailed reporting of many lesions can be time-consuming and is often avoided for cases with multiple lesions. For follow-up examinations, the 3D volume data can be either visually compared or the different scans can be aligned to each other (e.g., using a registration algorithm). However, an automatic registration might be error-prone if the local morphology has significantly changed, e.g., due to therapy or disease progression. Moreover, quantitative comparison between different lesions is only possible if these lesions have actually been reported.
Using CT imaging only, WO 2006/050102 proposed a 2D visualization technique, in the following referred to as “rib unfolding”, which improves the detectability of rib lesions. To this end, the centerline of each rib is extracted. Then, for each rib a 2D projection of its centerline (often also referred to as “curved planar reformation” (CPR)) is visualized. Coupling this technique with rib labelling, lesions and corresponding rib information (left/right rib, rib number) can be identified and used for reporting.
An example of this technique is shown in
However, using CT imaging only, rib lesions are often hard to detect as they may lack contrast to normal ribs and also may appear darker or brighter than healthy bone. As a consequence, each rib still requires carefully examination using the rib unfolded view in combination with traditional MPR images.
An object of the present invention is to address the above-described problems and to provide improvements upon the known devices and methods.
In general terms, one embodiment of the invention is a method for generating an enhanced image for display from medical imaging data of a subject, which includes the steps of: determining a feature of interest in the imaging data elongated in at least one dimension, obtaining a location of a line through the imaging data along the feature of interest, and generating from the imaging data a projection image along this line.
The method allows a simple, automatic and direct means for identifying any anomalies, such as lesions, along elongated features of interest in the imaging data, such as ribs.
Suitably, the projection image is a two-dimensional, maximum intensity projection image.
Preferably, the line is a centerline along the feature of interest. In embodiments, the feature of interest and the line follow an arcuate or non-linear path.
More preferably the feature of interest is a rib of the subject.
In an embodiment, the step of generating includes obtaining a series of sets of image data along the line; and using the series to generate the projection image.
Preferably, the series is composed of sets of planes of image data orthogonal to the line along the feature of interest.
The method can further include obtaining a segmentation of the feature of interest, and generating a plurality of projection images along the line, delineated by the segmentation of the feature of interest.
In embodiments, the projection image is generated from functional imaging data of the subject, and the method further includes displaying as the enhanced image a combination of: anatomical imaging data of the subject; and the projection image.
Preferably, the method further includes displaying in the enhanced image a projection image for the feature alongside the feature in the anatomical imaging data.
More preferably, the projection image generated from the functional imaging data is combined with anatomical imaging data along the line, and this combination is displayed alongside the feature in the anatomical imaging data.
Suitably, the method further includes repeating the generation of the projection image for a plurality of elongated features in the imaging data.
In one embodiment, the method further includes recording a finding of an anomaly for the feature of interest where a value for the projection image exceeds a threshold.
For example, the value may be an SUV or other intensity measure, signifying a hot-spot in that projection image.
Preferably, the anomaly is recorded as a lesion in the rib. In one embodiment, the method further includes, for a finding of a lesion in a given rib, recording for the finding one or more of: a side of the given rib, a number of the given rib, a segment of the rib or image volume in which the lesion is found, a distance to the spine of the subject, and an angle between a subject axis and the lesion finding.
Suitably, the segmentation is performed on functional imaging data of the subject, and the method further includes comparing the segmentation with the finding of an anomaly.
The invention also encompasses apparatus for generating an enhanced image for display from medical imaging data of a subject, having a processor configured to determine a feature of interest in the imaging data elongated in at least one dimension, obtain a location of a line through the imaging data along the feature of interest, and generate from the imaging data a projection image along the line, and a display device, with the processor being configured to cause the projection image to be displayed at the displaying device.
In embodiments of the invention, functional imaging data may be obtained from modalities such as PET or SPECT, and anatomical imaging data from modalities such as CT or MRI.
The above object also is achieved in accordance with the present invention by a non-transitory, computer-readable data storage medium encoded with programming instructions that, when the storage medium is loaded into a computerized control and evaluation system are an imaging apparatus, cause the computerized control and evaluation system to operate the imaging apparatus according to any of the above-described embodiment of the inventive method.
The above embodiments may be combined to provide further aspects and embodiments of the invention.
a and 5b are sets of diagrams illustrating combined rib-specific functional and anatomical image data according to embodiments of the invention.
a is an example of a visualization of the images of
b is a schematic diagram of
When the following terms are used herein, the accompanying definitions can be applied:
PET—Positron Emission Tomography
SPECT—Single-photon emission tomography
SUV—Standardised Uptake Value
FDG—F-18 fluorodeoxyglucose, a PET radiotracer
NaF—Sodium fluoride
Tc—Technetium
MIP—Maximum Intensity Projection (or Minimum Intensity Projection, usually denoted MinIP)
MRI—Magnetic Resonance Imaging
ROI/VOI—Region/volume of interest.
Segmentation—the process of labelling voxels as belonging to different classes. For example, in a medical image, each voxel may be labelled according to its tissue type.
CT—Computed Tomography
MPR—Multi-Planar Reconstruction
CAD Computer-aided diagnosis
Embodiments of the invention present methods for enhanced visualizations, automatic/interactive reporting, and follow-up examinations.
Embodiments of this invention enable the reading physician to easily and efficiently assess which ribs are affected by lesions and to interactively/automatically report the locations of each finding, even in the presence of multiple lesions. Furthermore, they enable the automatic identification of corresponding lesions among different studies from the same patient.
The following describes embodiments of the invention including a system and methods for efficient rib lesion reading, including methods for advanced visualization, detection, reporting, and follow-up. In particular it improves on previous 2D visualization techniques for rib analysis by PET/SPECT imaging and furthermore introduces new visualization techniques that benefit from the high contrast between lesions and non-lesions in molecular imaging and hence allow a fast assessment of all lesions along each rib. Once a lesion has been detected (manually, semi-automatically, or fully-automatically) its location can be automatically reported. Moreover, the location of each finding can be easily re-visited in follow-up studies, e.g., for comparison purposes. Overall, the described embodiments are expected to significantly improve the workflow of multimodal bone readings with focus on rib lesions.
The basics of the proposed first embodiment of the system are sketched in the flowchart in
The first input is one anatomical 3D image (202) of the chest that is used for extraction of the rib centerlines (204) and optionally a binary segmentation of the different ribs. Preferably this input is a CT image, which allows the extraction of rib centerlines/binary masks using for instance a method based on that described in Sowmya Ramakrishnan, Christopher Alvino, Leo Grady, and Atilla Kiraly, “Automatic Three-Dimensional Rib Centerline Extraction from CT Scans for Enhanced Visualization and Anatomical Context,” SPIE Medical Imaging 2011, 7962, pp. 79622×.
Note that once the rib centerlines are available, this input is not mandatory.
The second input consists of one or more functional images (206—PET, SPECT, MRI). If the modality is a type of molecular imaging (e.g., PET), different radiotracers such as 18F—NaF or 18F-FDG, may be used for imaging.
The second input shall be aligned to the first input either by means of simultaneous acquisition or by means of a registration algorithm (208). In case of follow-up studies, both inputs should be available for the different time-points. Note, however, that the image data does not need to be aligned across time-points.
Then, each rib is straightened (210) to create for instance a rib unfolding view similar to that displayed for CT only in
The transverse images can then be used as basis for an enhanced rib unfolding visualization (216), whereas the orthogonal images are used for rib-specific visualization (218). The different views are then combined with conventional viewing techniques into one system that allows the efficient navigation to potential lesions and synchronizes the different views for better assessment.
Moreover, each individual rib can be further subdivided into different segments based on some anatomical feature that can be utilized to automatically describe findings based on the rib of occurrence and the position along the rib. Finally, reported lesions can be linked to corresponding lesions in follow-up studies based on this information, for instance for quantitative analyses, without the necessity of aligning the input data across time-points.
In an initial embodiment, an enhanced visualization of the results is presented. By synchronizing the rib straightening and reformatting steps between different modalities, the existing system as shown in
Beyond that, it is also possible to create multiple successive functional reformattings and to compute a projection (e.g., MIP) along the resulting stack of computed 2D images. That is, the value of each overlay pixels then corresponds to the, e.g., maximum intensity value of a small range along the view of sight (a so-called “thin-slice MIP”). Furthermore, such projections can be limited to a binary mask of the rib structures, if existent. It is clear that all described overlay techniques can be also visualized without the corresponding functional images. Using this technique significantly improves the visibility of potential lesions.
A main embodiment of this invention provides a rib-specific visualization that allows the identification of ribs that are affected by a lesion at a glance. To this end, the orthogonal reformattings (shown in
a and 5b show 2D images of specific ribs (“rib patches”) with (
Again, such projections can be limited to a binary mask of the rib structures, if existent. Such “rib patches” could for instance be combined with a MIP of the PET data as shown in
a is an example visualisation for the complete set of rib-specific CT and MIP combinations shown in
As in
In the shown example, each rib centerline is visualized in a color that corresponds to the color of the frames of each individual rib patch. Additionally, rib labels (612) are displayed for each rib patch.
The navigation can be synchronized between all previously described views. Particularly, if selecting a position along a rib in the rib unfolding view or in the MIP, the corresponding rib patch can be highlighted. Vice versa, clicking on a rib patch, corresponding position can be highlighted in the different other views.
Moreover, a hot-spot detection or CAD algorithm can be applied to each resulting stack of orthogonal reformattings. The result of such an algorithm can be integrated into the described system by, e.g., displaying the number of detected lesions along each rib as additional label. Then, the user could for instance automatically “jump” between different lesions/hot-spots along a particular centerline by clicking multiple times onto the corresponding rib patch.
b is a schematic version of
Once a finding has been located either manually, semi-automatically, or fully automatically, its location can be reported based on the extracted centerline information. The report of each finding can automatically include the side/number of each rib and also the location along each rib. To this end, the ribs may be further divided into different segments, e.g., using landmarks such as midclavicular line, anterior/mid/posterior axillary line, and scapular line. Such segments can be either detected or approximated by the angle (706) between finding and spine (704) as exemplarily outlined in
To relocate a reported finding in a follow-up study, each finding can be (internally) stored using a more detailed description of its location. Instead of saving just the rib segment in which a lesion is located, its detailed position (e.g., distance to spine, angle between finding and spine, etc.) can be stored. With such information being available, it is straightforward to link lesions between different time-points even without any registration between the datasets. This also allows a detailed comparison between them, e.g., lesion sizes using a lesion segmentation algorithm.
All this information together could be exported into a detailed report or into some dedicated file format that can be interpreted by, e.g., a therapy planning system.
The rib-specific MIP/CT combinations presented alongside the whole body PET MIP (as shown in
Referring to
For example, a central processing unit 904 is able to receive data representative of medical scans via a port 905 which could be a reader for portable data storage media (e.g. CD-ROM); a direct link with apparatus such as a medical scanner (not shown) or a connection to a network.
Software applications loaded on memory 906 are executed to process the image data in random access memory 907.
The processor 904 in conjunction with the software can perform the steps such as determining a feature of interest in the imaging data elongated in at least one dimension, obtaining a location of a line through the imaging data along the feature of interest, and generating from the imaging data a projection image along the line.
A Man—Machine interface 908 typically includes a keyboard/mouse/screen combination (which allows user input such as initiation of applications) and a screen on which the results of executing the applications are displayed.
Although modifications and changes may be suggested by those skilled in the art, it is the intention of the inventor to embody within the patent warranted hereon all changes and modifications as reasonably and properly come within the scope of their contribution to the art.
| Number | Date | Country | Kind |
|---|---|---|---|
| 1117805.0 | Oct 2011 | GB | national |
