Combined rib and spine image processing for rapid evaluation of scans.

Abstract

A visualization scheme is generated based on automated segmentation to display the patient's anatomy. Data representing a three-dimensional diagnostic image is received. The ribs and vertebrae are segmented, and rib centerlines and vertebral center landmarks are detected and labeled. Three-dimensional position coordinates corresponding to the received data are mapped to a defined two-dimensional manifold plane, and then an interpolation technique is performed to complete the remaining three-dimensional position coordinates and deform the two-dimensional manifold plane. Image intensities at each coordinate of the deformed two-dimensional manifold plane are sampled from the three-dimensional diagnostic image space. From the sampled image intensities, a reformatted image is generated as manifold slices that display a continuous and linearized visual representation of the thorax. A three-dimensional visual representation of the thorax is generated from the multiple manifold slices.

Description

[Technical field]

[0001] The present invention relates to the field of medical imaging, and in particular to a method and apparatus for automatically generating a visual representation of ribs and vertebrae. [Background technology]

[0002] Medical professionals commonly use medical imaging to visualize a patient's anatomy, where the visual representation data is used to diagnose disease or injury. In trauma situations, medical personnel may rely on medical imaging data in the form of computed tomography (CT) scans for diagnostic purposes, e.g., detection of rib fractures. Summary of the Invention [Problem to be solved by the invention]

[0003] Viewing a chest or whole-body three-dimensional (3D) CT scan slice-by-slice is often a time-consuming process, especially when the target anatomical structure spans multiple slices (e.g., tracing 24 individual ribs).

[0004] Interpreting image scans, and more specifically scans in trauma or emergency departments, is a time-critical task that needs to be performed with high attention to avoid missing important findings. Often in these settings, imaging is based on whole-body scans, resulting in the generation of a large amount of image data that must be thoroughly examined. The ribs and spine are particularly important structures to be evaluated. These are repetitive structures that require a significant amount of time to examine, and in the case of the spine, these structures are important from a neurological point of view.

[0005] Visualizing a patient's anatomy and certain critical structures in an intuitive manner could simplify the time-consuming examinations currently required. Several approaches have been proposed to simplify the visualization and evaluation of a patient's anatomy and critical structures, particularly those aimed at visualization of the thorax. While offering some advantages, current approaches suffer from clear fundamental limitations.

[0006] One well-known visualization scheme is the "fillet view" or "fishbone view", which involves segmenting the ribs (e.g., using a deep convolutional neural network) and then labeling pairs of ribs in the field of view using a centerline extractor. Each rib is sampled along its trace, which allows visualization of each and every rib in a normalized and straightened manner (curved planar reformatted). This view allows medical professionals to accurately inspect rib centerlines in a normalized view (reformatted view) where every rib is straightened and placed in a unique position on the inspection canvas.

[0007] However, this type of view has several drawbacks. One drawback with this type of view is that the nature of processing each rib independently leads to discontinuities between ribs, which can lead to image artifacts from adjacent ribs appearing in the rib shape. In addition, this view is limited to only the ribs, and adding context information such as the spine requires additional steps and image processing that must be done separately.

[0008] Another visualization scheme is the visceral cavity view. In this view, a segmentation algorithm (e.g., using a model-based approach) is applied to segment the inside of the rib cage in terms of a deformation of a cylindrical manifold. Once segmented, the manifold can be unwrapped and a maximum intensity projection (MIP) near the surface can be calculated. This view allows the user to inspect the entire rib cage in terms of a continuous visual representation on an inspection canvas.

[0009] One drawback of this type of view is that relative rib lengths are not maintained. Due to the nature of unwrapping the cylindrical manifold, it is not possible to visualize the correct rib lengths. For example, the first rib appears too long relative to the other ribs. In addition, due to the nature of the MIP visualization, microfractures may not be detectable. For example, small rib fractures may be hidden and undetectable in the generated view with the MIP visualization. Another drawback is that this view adds large unrealistic distortions (wavy ribs) that limit the clinical reliability of the generated visceral cavity view.

[0010] Therefore, there is a need for an innovative visualization scheme that combines the advantages of the fillet view and the visceral cavity view without the drawbacks associated with these views.

[0011] It is an object of the present invention to provide a technique for generating new manifold views that overcomes the shortcomings of existing visualization schemes. [Means for solving the problem]

[0012] The technology can be applied to several imaging systems, including CT, CT arms, single photon emission computed tomography CT (SPECT-CT), magnetic resonance CT (MR-CT), positron emission tomography CT (PET-CT), and magnetic resonance imaging (MRI) systems.

[0013] According to a first aspect of the present invention, there is provided an image processing device having a memory configured to store computer executable instructions and at least one processor configured to execute the computer executable instructions to cause the image processing device to receive data representing a three-dimensional diagnostic image including ribs and a vertebrae of a subject, the computer executable instructions causing the image processing device to perform the steps of segmenting the ribs and vertebrae according to the received data representing the three-dimensional diagnostic image, detecting and labeling rib centerlines from the rib segmentation, detecting and labeling vertebral center landmarks from the vertebrae segmentation, and defining a two-dimensional manifold plane representing a visualization canvas for displaying the reformatted image. The computer executable instructions cause the image processor to perform the steps of: mapping 3D positions of rib centerline and vertebral center landmarks corresponding to received data representing a 3D diagnostic image to 2D positions on a defined 2D plane; interpolating missing 3D position coordinates on the defined 2D manifold plane, including deforming the 2D manifold plane so that it aligns with the detected rib centerline and vertebral center landmarks; and sampling image intensities at each coordinate of the deformed 2D manifold plane from the 3D diagnostic image space. From the sampled image intensities, a reformatted image is generated as manifold slices, which display a continuous and linearized visual representation of the rib cage and spine. The deformed 2D manifold plane is shifted along its normal direction and the sampling is repeated to generate a stack of manifold slices that cover the complete 3D visual representation of the rib cage. The 3D visual representation of the rib cage and vertebral centers is generated as a stack of manifolds that cover the entire rib cage.

[0014] In a second aspect of the invention, there is provided a method of processing a three-dimensional image, the method comprising the steps of receiving data representing a three-dimensional diagnostic image including ribs and a vertebrae of a subject, segmenting the ribs and vertebrae according to the received data representing the three-dimensional diagnostic image, detecting and labeling rib centerlines from the rib segmentation, detecting and labeling vertebral center landmarks from the vertebral segmentation, and defining a two-dimensional manifold plane representing a visualization canvas for displaying the reformatted image. The method includes the steps of mapping 3D positions of rib centerline and vertebral center landmarks corresponding to received data representing a 3D diagnostic image to 2D positions on a defined 2D plane, interpolating missing 3D position coordinates on a defined 2D manifold plane, including deforming the 2D manifold so that the 2D manifold plane aligns with the detected rib centerline and vertebral center landmarks, and sampling image intensities at each coordinate of the deformed 2D manifold plane from the 3D diagnostic image space. A reformatted image is generated from the sampled image intensities as manifold slices, the manifold slices displaying a continuous and linearized visual representation of the rib cage and spine. The deformed 2D manifold plane is shifted along its normal direction and the sampling is repeated to generate a stack of manifold slices covering the complete 3D visual representation of the rib cage. The 3D visual representation of the rib cage and vertebral centers is generated as a stack of manifolds covering the entire rib cage.

[0015] In a third aspect of the invention, a non-transitory computer readable medium having stored thereon instructions for causing a processing circuit to execute a process comprising the steps of receiving a three-dimensional diagnostic image including ribs and a vertebrae of a subject, segmenting the ribs and the vertebrae according to the received data representing the three-dimensional diagnostic image, detecting and labeling rib centerlines from the rib segmentation, detecting and labeling vertebral central landmarks from the vertebral segmentation, and defining a two-dimensional manifold plane representing a visualization canvas for displaying the reformatted image, the two-dimensional manifold plane being deformed to align with the detected rib centerlines and vertebral central landmarks. The process further includes mapping each of the three-dimensional locations of the rib centerline and vertebral center landmarks corresponding to the received data representing the three-dimensional diagnostic image to a two-dimensional location on the defined two-dimensional plane, and interpolating missing three-dimensional location coordinates on the defined two-dimensional manifold plane, including deforming the two-dimensional manifold plane so that it aligns with the detected rib centerline and vertebral center landmarks. Image intensities at each coordinate of the deformed two-dimensional manifold plane are sampled from the three-dimensional diagnostic image space. From the sampled image intensities, reformatted images are generated as manifold slices, which display a continuous and linearized visual representation of the rib cage and spine. The deformed two-dimensional manifold plane is shifted along its normal direction and the sampling is repeated to generate a stack of manifold slices that cover the complete three-dimensional visual representation of the rib cage. The three-dimensional visual representation of the rib cage and vertebral centers is generated as a stack of manifolds that cover the entire rib cage.

[0016] In a preferred embodiment, the segmentation of the ribs and vertebrae is performed using machine learning or deep learning techniques, including at least one of neural networks, logistic regression, random forests, nearest neighbors, and cluster or multivariate analysis. A deep learning based segmentation approach is applied to segment the ribs and vertebrae, then extract rib centerlines and detect vertebral center landmarks. As a result, all rib pairs and vertebral centers of gravity are accurately detected and labeled in an automated manner.

[0017] A 2D manifold plane is defined that is used to display a visual representation canvas for displaying the reformatted image. Each 3D location of the detected rib centerline and vertebral center landmarks is mapped to a 2D location on the visualization canvas that visualizes the unfolded ribs. That is, a coordinate correspondence from 2D space to 3D space is defined.

[0018] Once the correspondence from 2D to 3D coordinates has been established, the correspondences from the remaining pixels in the 2D space still lag behind. In one aspect of the invention, in those regions where this information is not yet available, an interpolation scheme is implemented that interpolates the x / y / z coordinate correspondence from 3D space to 2D space.

[0019] In a preferred embodiment, the interpolation is performed using thin plate spline techniques. Thin plate splines are a good choice for the interpolation technique because they model the deformation of an originally thin two-dimensional metal plate. The image intensities at each coordinate of the deformed two-dimensional manifold plane are sampled from the three-dimensional diagnostic image space. Reformatted images are then generated from the sampled image intensities as manifold slices, which display a continuous and linearized visual representation of the rib cage and spine.

[0020] In one aspect of the invention, the sampled image intensities at each location of the deformed two-dimensional manifold plane correspond to at least one of a region of tissue between the ribs and a region of tissue adjacent to the ribs.

[0021] In one aspect of the invention, a stack of interpolated 2D multiplanar reconstructions (MPRs) is generated by calculating the normal direction of each point on the 2D plane and shifting or moving the manifold surface multiple times inward and outward of the manifold plane. A 3D image of the rib cage and spine is then generated based on the stack of 2D MPRs. [Brief description of the drawings]

[0022] [Figure 1] 1 is a diagram of an exemplary imaging system. [Diagram 2] FIG. 2 shows a "fillet view" or "fishbone view" of the rib cage. [Diagram 3] Diagram showing the visceral cavities of the thorax. [Figure 4] 1 is a flowchart illustrating a method for generating a composite rib and spine view, according to some embodiments. [Diagram 5] 1 illustrates a mapping from three-dimensional space to two-dimensional space, according to some embodiments. [Figure 6] FIG. 1 illustrates a processing workflow according to some embodiments. [Figure 7] FIG. 2 illustrates a three-dimensional view of the ribs and spine according to some embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] 1 illustrates an imaging system 100, such as a computed tomography (CT) imaging system. The imaging system 100 generally includes a stationary gantry 102 and a rotating gantry 104. The rotating gantry 104 is rotatably supported by the stationary gantry 102 and rotates relative to the stationary gantry 102 about an examination region 106 about a longitudinal axis, or z-axis.

[0024] A patient support 112, such as a couch, supports a subject, such as a human patient, in the examination region 106. The patient support 112 is configured to move the subject for loading, scanning, and unloading.

[0025] A radiation source 108, such as an x-ray tube, is rotatably supported by the rotating gantry 104. The radiation source 108 rotates with the rotating gantry 104 and emits radiation that traverses an examination region 106.

[0026] A radiation sensitive detector array 110 forms an angular arc across the examination region 106 on an opposite side to the radiation source 108. The detector array 110 extends along the z-axis direction and has one or more rows of detectors that detect radiation traversing the examination region 106 and generate projection data indicative of the detected radiation.

[0027] A general purpose computing system or computer serves as an operator console 114 and has input devices 116, such as a mouse, keyboard, and / or the like, and output devices 120, such as a display monitor, filmer, etc. The console 114 allows an operator to control the operation of the system 100, including control of an image processor 118, which can receive data representing three-dimensional (3D) diagnostic images produced by the imaging system 100 and generate reformatted images that display a continuous, linearized visual representation of the rib cage and spine. The reformatted images can be output to the output device 120 for review by medical personnel.

[0028] It will be appreciated that the processing of the image processor 118 may be realized via a processor and / or processing circuitry executing computer-readable instructions, such as executable code, encoded on or embedded on a computer-readable storage medium, such as physical memory and other non-transitory media. Additionally or alternatively, the processor(s) and / or processing circuitry may execute computer-readable instructions propagated by carrier waves, signals, and other transitory (or non-transitory) media.

[0029] Figure 2 shows a view of the rib cage in the "fillet view" or "fishbone view" visualization scheme, which allows medical professionals to accurately inspect rib centerlines in a normalized (reformatted) view, where all ribs are seen to be straightened and placed in a unique position on the inspection canvas.

[0030] One drawback of this view is that the nature of processing each rib independently can result in discontinuities between ribs that can result in image artifacts appearing in the rib shapes from adjacent ribs as identified at 202. Another drawback is that the field of view is limited to just the ribs.

[0031] FIG. 3 shows a diagram of the visceral cavity. In this visual scheme, a segmentation algorithm using, for example, a model-based approach is applied to segment the inside of the rib cage on a deformation of a cylindrical manifold. Once segmented, the manifold can be unwrapped and a maximum intensity projection (MIP) close to the surface can be calculated.

[0032] This view allows the user to inspect the entire rib cage with respect to a continuous visual representation on the inspection canvas. One drawback of the visceral cavity view seen in FIG. 3 is that relative rib lengths are not maintained. For example, the first rib appears very long relative to the other ribs. In addition, due to the nature of MIP visualization, microfractures may not be detectable. For example, MIP visualization may make small rib fractures invisible and undetectable in the generated view. Another drawback is that this view adds significant unrealistic distortions (wavy ribs) that limit clinical confidence in the generated visceral cavity view.

[0033] 4 is a flow chart illustrating a method for generating a reformatted medical image.The type of medical imaging performed may be, for example, a CT scan to generate a three-dimensional CT image of a human patient.

[0034] At block 402, a three-dimensional diagnostic image is received. As an example, the three-dimensional diagnostic image may be a CT imaging system scan. At block 404, the ribs and spine of the scanned patient are automatically segmented. The segmentation may be performed using machine learning or deep learning techniques. As an example, Convolutional Neural Networks (CNN) are a deep learning technique that can be used to segment the ribs and spine. CNN architectures such as U-Net and V-Net can also be adapted for the segmentation task. Deep learning-based training using datasets from hundreds of scans can be used to develop algorithms that are shown to achieve fast and accurate segmentation of the ribs and spine.

[0035] As part of the process, in block 406, ribs are detected, rib centerlines are extracted and labeled, and in block 408, vertebral center landmarks are detected and labeled. In block 410, a two-dimensional (2D) manifold plane is defined. The 2D manifold plane is defined to represent the visualization canvas of the reformatted images that will be generated later. The 2D manifold plane is defined such that the visualization canvas of the reformatted images is a standardized visualization of the rib cage. Thus, the standardized visual representation of the rib cage can be used to compare with follow-up scans without the need for image registration, since the standardized views are already registered. In other words, a one-to-one mapping between coordinates in one space and coordinates in another space is determined such that points in the two spaces correspond to the same anatomical points.

[0036] In block 412, each three-dimensional (3D) location of the rib centerline and vertebral center landmarks corresponding to the received data, the scan image data, representing a three-dimensional diagnostic image, is mapped to a 2D location on the defined two-dimensional manifold plane. Since the positions of the ribs and vertebrae are known in the 3D space, correlations for all the rib centerlines and all the vertebral center landmarks can be established. That is, the coordinate correspondence from the 3D space to the 2D space can be defined as a mapping function. Once the 3D to 2D correspondence is established or set up, there may be missing correspondences or correlations from the remaining pixels in the defined 2D space.

[0037] Therefore, for regions where this information is missing, an interpolation scheme is implemented in block 414 to interpolate x / y / z coordinate correspondences from the 3D space to the prescribed 2D space. This information may include tissue regions between the ribs and adjacent to the ribs. This information may also include portions of the lungs, which may be crucial when pneumothorax and hemothorax coexist. The interpolation scheme deforms the prescribed 2D manifold so that it aligns with the detected rib centerlines and vertebral center landmarks. Thin plate splines are a good choice to use as an interpolation scheme since they inherently model the deformation of thin 2D metal plates. Elastic body splines (EBS) are another approach to mapping or coordinate transformation that is suitable for medical image analysis.

[0038] In block 416, image intensities from the 3D diagnostic image space are sampled at each coordinate of the transformed 2D manifold plane. Sampling image intensities from the 3D diagnostic image space at each coordinate of the transformed 2D manifold plane ultimately unwraps the rib cage and results in a reformatted image. The sampling of image intensities can be based on a coordinate correspondence from 3D space to 2D space as a mapping function. That is, sampling is performed according to a coordinate mapping function that corresponds to the mapped and interpolated positions.

[0039] At block 418, a reformatted image is generated from the sampled image intensities as manifold slices. The reformatted image generated as manifold slices displays a continuous and straightened visual representation of the rib cage and spine. The reformatted image combines the advantages of both the fillet view and the visceral cavity view to display a normalized and straightened rib view while maintaining relative rib lengths, and the reformatted image avoids the disadvantages of both the fillet view and the visceral cavity view, including discontinuities between ribs, imaging artifacts appearing from adjacent rib positions, inaccurate rib lengths, and non-visible fractures and distortions due to MIP. The reformatted image generated at block 418 can be output to a medical professional for diagnostic purposes.

[0040] To facilitate further examination and diagnosis, a stack of 2D manifold slices can be generated. In block 420, the 2D manifold plane is shifted along its normal direction towards the inside and outside of the manifold plane, and in block 422, sampling is repeated for each incremental shift along the normal direction to generate a stack of manifold slices. The stack of manifold slices covers a complete 3D visual representation of the rib cage and spine.

[0041] In block 424, a 3D visual representation covering the complete rib cage and vertebral centers is generated as a stack of manifolds covering the complete rib cage. The stack of manifold slices generated in block 422 can be a stack of interpolated 2D multiplanar reconstructions (MPRs), and in block 424, a 3D image of the rib cage and spine can be generated based on the stack of 2D MPRs. The generated 3D visual representation covering the entire rib cage and vertebral centers allows for inclusion of all ribs in their entirety. This allows medical professionals to easily browse the rib cage to quickly visually inspect the rib cage and look for rib cage and spinal fractures.

[0042] FIG. 5 illustrates the mapping from 3D space to 2D space. The positions of the ribs and vertebrae are known in 3D space. Each position of the rib centerline and vertebral center landmarks in 3D space is mapped to a corresponding position in the prescribed 2D manifold plane. The prescribed 2D manifold plane is the visualization canvas for the 2D manifold image that will be generated later. In essence, the ribs are traced and each rib centerline and vertebral center landmark is mapped to a pixel location where it appears in the 2D manifold. For ease of illustration, only three rib and vertebral center landmarks are shown in FIG. 5. However, it should be understood that all rib and vertebral center landmarks that correspond to the rib and vertebral center landmarks known in 3D space are mapped to a position in the prescribed 2D manifold plane.

[0043] FIG. 6 illustrates a processing workflow according to some embodiments. At 602, rib centerlines and vertebral center landmarks are detected and labeled according to an automated segmentation scheme. The image segmentation scheme can be based on machine learning or deep learning techniques. A 2D manifold is defined that represents the resulting reformatted view. At 604, an interpolation scheme deforms the 2D manifold to align with the detected rib centerlines and vertebral center landmarks. Thin plate splines or any suitable interpolation scheme can be used. As shown at 604, the defined 2D manifold is deformed to align with the detected rib centerlines and vertebral center landmarks that are mapped from the 3D space to the 2D space of the 2D deformation manifold.

[0044] Image intensities from a 3D image space, for example a 3D image from a CT scan, are sampled at each coordinate of the deformed 2D manifold to complete or fill in the missing 3D coordinates, as shown at 606. The missing 3D coordinates may correspond to regions of tissue between the ribs and regions of tissue adjacent to the ribs, or any combination of regions of tissue between the ribs and regions of tissue adjacent to the ribs. This allows the deformed 2D manifold to be unfolded.

[0045] The unfolded 2D manifold is shown at 608. The new manifold view, shown as a reformatted image at 608, combines the advantages of both the fillet view and the visceral cavity view while avoiding the disadvantages of both. At 608, it can be seen that a normalized and straightened view of the ribs is displayed while maintaining relative rib lengths. Inter-rib discontinuities, image artifacts appearing from adjacent rib positions, inaccurate rib lengths, and unseen fractures and distortions due to MIP are avoided.

[0046] FIG. 7 shows a 3D view of the ribs and spine, according to some embodiments. A 3D visual representation of the rib cage and vertebral centers can be generated by shifting or moving the deformed 2D manifold plane along its normal direction multiple times toward the inside and outside of the manifold plane, and repeatedly sampling to generate a stack of 2D manifold slices. This results in a 3D image as a stack of manifolds covering the complete rib cage. FIG. 7 shows that the resulting 3D image allows for the inclusion of all the ribs in their entirety.

[0047] The stack or manifold slices can be a stack of interpolated 2D multiplanar reconstructions (MPRs) and a 3D image of the rib cage and spine can be generated based on the stack of 2D MPRs. The generated 3D rendering is shown at 702. The view at 704 has been rotated approximately 30 degrees and the view at 706 has been rotated slightly less than 90 degrees.

[0048] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive, and the invention is not limited to the disclosed embodiments.

[0049] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.

[0050] In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality.

[0051] A single processor, device or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0052] Operations such as obtaining, determining, acquiring, outputting, providing, storing, calculating, simulating, receiving, alerting, and stopping can be realized as program code means of a computer program and / or as dedicated hardware. The computer program can be stored and / or distributed on a suitable medium, such as an optical storage medium or a solid-state medium, provided together with or as part of other hardware, but can also be distributed in other forms, such as over the Internet or other wired or wireless telecommunication systems.

Claims

1. An image processing device, a memory storing computer-executable instructions; at least one processor that executes computer-executable instructions; the computer-executable instructions causing the image processing device to: receiving data representing a three-dimensional diagnostic image including the ribs and spine of a subject; segmenting ribs and vertebrae according to the received data representing the three-dimensional diagnostic image; Detecting and labeling rib centerlines from the rib segmentation; detecting and labeling vertebral body center landmarks from the spinal segmentation; defining a two-dimensional manifold plane representing a visualization canvas for displaying the reformatted image; mapping the three-dimensional positions of the rib centerlines and the vertebral center landmarks corresponding to the received data representing the three-dimensional diagnostic image to two-dimensional positions on the defined two-dimensional manifold plane; interpolating missing three-dimensional position coordinates on the defined two-dimensional manifold plane, the step including deforming the two-dimensional manifold plane so that the two-dimensional manifold plane aligns with the detected rib centerlines and vertebral center landmarks; sampling image intensities at each coordinate of the deformed two-dimensional manifold plane from the three-dimensional diagnostic image space; generating the reformatted image as manifold slices from the sampled image intensities, the manifold slices displaying a continuous and linearized visual representation of the rib cage and spine; shifting the deformed 2D manifold plane along its normal direction and repeating the sampling to generate a stack of manifold slices that covers the complete 3D visual representation of the rib cage and spine; generating a three-dimensional visual representation of the thorax and vertebral centers as a stack of manifolds covering the entire thorax; An image processing device that executes the above.

2. The image processing device of claim 1 , wherein the segmentation of the ribs and the vertebrae is performed by machine learning or deep learning techniques including at least one of neural networks, logistic regression, random forests, nearest neighbors, and cluster or multivariate analysis.

3. The image processing apparatus of claim 1 , wherein the interpolation is performed by a thin-plate spline technique.

4. The image processing device of claim 1 , wherein the processor further executes the computer-executable instructions to cause the image processing device to calculate a normal direction for each point on the two-dimensional manifold plane.

5. The image processing apparatus of claim 1 , wherein the generated stack of manifold slices is a stack of interpolated two-dimensional multiplanar reconstructions.

6. The image processing device of claim 5 , wherein the processor is further configured to execute the computer-executable instructions to generate a three-dimensional image of the rib cage and spine based on the stack of two-dimensional multi-planar reconstructions.

7. The image processing device of claim 1 , wherein the sampled image intensities at each location of the deformed two-dimensional manifold plane correspond to at least one of a region of tissue between the ribs and a region of tissue adjacent to the ribs.

8. 1. A computer-implemented method for processing three-dimensional images, comprising: receiving data representing a three-dimensional diagnostic image including the ribs and spine of a subject; segmenting ribs and vertebrae according to the received data representing the three-dimensional diagnostic image; Detecting and labeling rib centerlines from the rib segmentation; Detecting and labeling vertebral body center landmarks from the spine segmentation; defining a two-dimensional manifold plane representing a visualization canvas for displaying the reformatted image; mapping the three-dimensional positions of the rib centerlines and the vertebral center landmarks corresponding to the received data representing the three-dimensional diagnostic image to two-dimensional positions on the defined two-dimensional plane; interpolating missing three-dimensional position coordinates on the defined two-dimensional manifold plane, the step including deforming the two-dimensional manifold plane so that the two-dimensional manifold plane aligns with the detected rib centerlines and vertebral center landmarks; sampling image intensities from the three-dimensional diagnostic image space at each coordinate in the transformed two-dimensional manifold plane; generating reformatted images from the sampled image intensities as manifold slices, the manifold slices displaying a continuous and linearized visual representation of the rib cage and spine; shifting the deformed 2D manifold plane along its normal direction and repeating the sampling to generate a stack of manifold slices that cover a complete 3D visual representation of the rib cage and spine, whereby the 3D visual representation of the rib cage and vertebral centers is generated as a stack of manifolds that cover the complete rib cage; A method having the following.

9. 9. The method of claim 8, wherein the segmentation is performed using machine learning or deep learning techniques, the machine learning or deep learning techniques including at least one of neural networks, logistic regression, random forests, nearest neighbors, and cluster or multivariate analysis.

10. The method of claim 8 , wherein the interpolation is performed by a thin-plate spline technique.

11. The method of claim 8 , further comprising the step of calculating the normal direction for each point on the two-dimensional plane.

12. The method of claim 8 , wherein the generated stack of manifold slices is a stack of interpolated two-dimensional multiplanar reconstructions.

13. The method of claim 12 further comprising generating a three-dimensional image of the rib cage and spine based on the stack of interpolated two-dimensional MPR images.

14. 9. The method of claim 8, wherein the sampled image intensities at each location of the deformed two-dimensional manifold plane correspond to at least one of a region of tissue between the ribs and a region of tissue adjacent to the ribs.

15. A non-transitory computer-readable medium having stored thereon instructions for causing a processing circuit to perform a process, the process comprising: receiving data representing a three-dimensional diagnostic image including the ribs and spine of a subject; segmenting ribs and vertebrae according to the received data representing the three-dimensional diagnostic image; Detecting and labeling rib centerlines from the rib segmentation; Detecting and labeling vertebral body center landmarks from the spine segmentation; defining a two-dimensional manifold plane representing a visualization canvas for displaying the reformatted image, the two-dimensional manifold plane being deformed to align it with the detected rib centerlines and vertebral center landmarks; mapping the three-dimensional positions of the rib centerlines and vertebral center landmarks corresponding to the received data representing the three-dimensional diagnostic image to two-dimensional positions on the defined two-dimensional plane; interpolating missing three-dimensional position coordinates on the defined two-dimensional manifold plane, the step including deforming the two-dimensional manifold plane so that the two-dimensional manifold plane aligns with the detected rib centerlines and vertebral center landmarks; sampling image intensities from the three-dimensional diagnostic image space at each coordinate in the transformed two-dimensional manifold plane; generating the reformatted image as manifold slices from the sampled image intensities, the manifold slices displaying a continuous and linearized visual representation of the rib cage and spine; shifting the deformed 2D manifold plane along its normal direction and repeating the sampling to generate a stack of manifold slices that covers a complete 3D visual representation of the entire rib cage; generating a three-dimensional visual representation of the thorax and vertebral centers as a stack of manifolds covering the entire thorax; 1. A computer-readable medium having:

Images