Medical image processing apparatus and medical image processing method

Abstract

To facilitate visualization of thin and / or small spatial features.SOLUTION: A medical image processing apparatus according to an embodiment comprises a processing circuit. The processing circuit receives a slab including a plurality of samples specified by a camera model, determines whether or not one or more first samples out of the plurality of samples constitute a part of an anatomical region of interest, and forms an image by projecting the slab onto an image plane along a line-of-sight direction. When determining that the one or more first samples constitute a part of the anatomical region of interest, the processing circuit projects the one or more first samples in a first projection mode, and projects one or more second samples out of the plurality of samples, which are not a part of the anatomical region of interest, in a second projection mode.SELECTED DRAWING: Figure 2A

Description

[Technical Field]

[0001] FIELD OF THE INVENTION The embodiments disclosed herein generally relate to medical imaging devices and methods. [Background technology]

[0002] Volume rendering is the process of computing two-dimensional (2D) images of three-dimensional (3D) objects. Volume rendering is applied, for example, to the rendering of medical volume data obtained by scanning the human body with computed tomography (CT) and other X-ray scanners, nuclear magnetic resonance scanners, ultrasound scanners, and other medical scanners.

[0003] Volume data consists of multiple voxels arranged in a 3D grid. Each voxel has a corresponding voxel value. The voxel value represents a measurement of a physical parameter. For example, in a CT scan, the voxel value represents the voxel's opacity to X-rays, i.e., its X-ray stopping power. X-ray stopping power is measured in Hounsfield units (HU), which correlate closely with density (unit mass of volume).

[0004] The voxels of volume data collected by medical scanners are often collected on a Cartesian grid, i.e., the data points are aligned along three orthogonal axes that define the volume space.

[0005] A 2D image is composed of multiple pixels arranged in a 2D grid. Visual space can be defined by three orthogonal axes, X, Y, and Z, with a common origin in one corner of the image. The X and Y axes extend through the image plane and are aligned with the 2D grid of pixels. The Z axis is perpendicular to the image plane and parallel to the line of sight.

[0006] Images are generated from volume data using conventional slab multi-planar reformatting (MPR). In this method, MPR data is generated by taking coordinates in view space, transforming them into volume space, and sampling the volume data using an interpolation method, such as trilinear interpolation, to generate MPR data values ​​at discrete view space coordinates. MPR data values ​​are also referred to as MPR samples. MPR slices are formed by performing the above process for multiple coordinates in the image plane. By repeating this process along a line of sight perpendicular to the image plane, multiple MPR slices can be identified and projected to form an MPR slab. Thus, an MPR slab consists of a series of MPR slices parallel to the image plane and aligned at different positions along the line of sight.

[0007] The 2D image is formed by projecting (folding) the MPR slab onto the image plane along the line of sight, using a projection algorithm. Examples of projection algorithms used to project the MPR slab include Maximum Intensity Projection (MIP), Minimum Intensity Projection (MinIP), and Average Intensity Projection (AveIP).

[0008] For example, the MIP algorithm is based on determining, for each pixel, the maximum voxel value along the Z axis within the MPR slab for the pixel's corresponding XY coordinate. MIP is a ray-casting method. For each pixel in the image, a virtual ray is cast into the volume data parallel to the line of sight. The image data for each pixel is the maximum voxel value hit by the ray as it traverses the MPR slab. The MinIP algorithm uses the minimum voxel value hit by the ray as its image data, rather than the maximum value. The AveIP method averages voxel data values ​​sampled from a portion of the ray traversing the slab to generate an overall value.

[0009] When utilizing these projection algorithms, it can be difficult to visualize thin and / or small spatial features that define or are part of an anatomical region of interest. For example, it can be difficult to visualize thin fractures in bone within a projection image using these projection algorithms. As a result, it can be difficult to determine the size and / or extent of a fracture. [Prior art documents] [Patent documents]

[0010] [Patent Document 1] US Patent Application Publication No. 2017 / 0262978 Summary of the Invention [Problem to be solved by the invention]

[0011] One of the problems to be solved by the embodiments disclosed in this specification and the drawings is to facilitate visualization of thin and / or small spatial features. However, the problems to be solved by the embodiments disclosed in this specification and the drawings are not limited to the above problem. Problems corresponding to the effects of each configuration shown in the embodiments described below can also be positioned as other problems. [Means for solving the problem]

[0012] According to an embodiment, a medical image processing apparatus includes a processing circuit that receives a slab including a plurality of samples identified by a camera model, determines whether one or more first samples of the plurality of samples form part of an anatomical region of interest, and forms an image by projecting the slab onto an image plane along a line of sight. If the processing circuit determines that the one or more first samples form part of the anatomical region of interest, the processing circuit projects the one or more first samples in a first projection mode and projects one or more second samples of the plurality of samples that are not part of the anatomical region of interest in a second projection mode. [Brief explanation of the drawings]

[0013] Embodiments will now be described, by way of non-limiting example, with reference to the accompanying drawings, in which: [Figure 1] FIG. 1 is a schematic diagram of a medical image processing apparatus according to an embodiment. [Figure 2A] FIG. 2A is a flowchart showing an overview of a process according to one embodiment. [Figure 2B] FIG. 2B is a flow chart illustrating other steps that are part of the process of FIG. 2A. [Figure 3] FIG. 3 is a schematic diagram illustrating an example of candidate regions detected in one of the steps of FIG. 2B. [Figure 4A] FIG. 4A is a schematic diagram illustrating a sphere generated as part of the sphere filtering method that constitutes one of the steps of FIG. 2B. [Figure 4B] FIG. 4B is a schematic diagram of a portion of the sphere shown in FIG. 4A, including multiple predicted gradient vectors. [Figure 5A] FIG. 5A shows a two-dimensional image of an anatomical region of interest generated by projecting an MPR slab using maximum intensity projection mode. [Figure 5B] FIG. 5B shows a two-dimensional image of an anatomical region of interest generated by projecting an MPR slab using minimum intensity projection mode. [Figure 5C] FIG. 5C shows a two-dimensional image of an anatomical region of interest generated using the processes of FIGS. 2A and 2B. [Figure 6A] FIG. 6A shows a two-dimensional image of an anatomical region of interest generated by projecting an MPR slab in mean projection mode. [Figure 6B] FIG. 6B shows a two-dimensional image of an anatomical region of interest generated using the processes of FIGS. 2A and 2B. [Figure 7] FIG. 7 shows a two-dimensional image of a mask defined by pixels associated with an anatomical region of interest. [Figure 8] FIG. 8 shows another two-dimensional image of an anatomical region of interest generated using the processes of FIGS. 2A and 2B. [Figure 9] FIG. 9 shows a two-dimensional image of another anatomical region of interest generated by projecting an MPR slab using maximum intensity projection mode. [Figure 10] FIG. 10 shows a two-dimensional image of the anatomical region of interest of FIG. 9 produced using the processes of FIGS. 2A and 2B. DETAILED DESCRIPTION OF THE INVENTION

[0014] According to an embodiment, a medical image processing apparatus includes a processing circuit that receives a slab including a plurality of samples identified by a camera model, determines whether one or more first samples of the plurality of samples are part of an anatomical region of interest, and forms an image by projecting the slab onto an image plane along a line of sight. If the processing circuit determines that the one or more first samples are part of the anatomical region of interest, the processing circuit projects the one or more first samples in a first projection mode and projects one or more second samples of the plurality of samples that are not part of the anatomical region of interest in a second projection mode.

[0015] An exemplary medical image processing method includes receiving a slab including a plurality of samples identified by a camera model, determining whether one or more first samples of the plurality of samples form part of an anatomical region of interest, and forming an image by projecting the slab onto an image plane along a line of sight, the method including, if the one or more first samples are determined to form part of the anatomical region of interest, projecting the one or more first samples in a first projection mode and projecting one or more second samples of the plurality of samples that are not part of the anatomical region of interest in a second projection mode.

[0016] 1 shows an outline of a medical image processing apparatus 10 according to one embodiment. The medical image processing apparatus 10 includes a computing device 12. The computing device 12 is a personal computer, a workstation, or the like. In this embodiment, the computing device 12 is connected to a scanner 14 via, for example, a data storage unit 16, but in another embodiment, the medical image processing apparatus 10 does not need to be connected or coupled to a scanner.

[0017] The medical imaging system 10 also includes one or more display screens 18 and one or more input devices, such as a computer keyboard, mouse, trackball, or the like.

[0018] In this embodiment, the scanner 14 is a computed tomography (CT) scanner, although in other embodiments the scanner may include a medical scanner, such as a nuclear magnetic resonance scanner, an ultrasound scanner, or other medical scanner. The scanner 14 generates image data representative of an anatomical region of a patient or other subject.

[0019] In this embodiment, the image data sets acquired by scanner 14 are stored in data store 16 and provided to computing device 12. In alternative embodiments, the image data sets may be provided from a remote data store (not shown). Data store 30 or the remote data store may comprise any suitable type of memory.

[0020] The computing device 12 includes processing circuitry 22 for data processing. The processing circuitry 22 includes a central processing unit (CPU) and a graphics processing unit (GPU). The processing circuitry 22 provides processing resources for automatic or semi-automatic processing of medical image data sets. In alternative embodiments, the data to be processed may include any image data other than medical image data.

[0021] In this embodiment, the computing device 12 includes image processing circuitry 24 for generating a slab containing multiple samples identified by the camera model. The slab is generated according to the multi-planar reformatting (MPR) method described above, but other reconstruction methods may also be used. For example, the slab may be generated based on curved planar reformatting (CPR), the reconstruction method described in U.S. Patent Application Publication No. 2017 / 0262978, or other reconstruction methods. The image processing circuitry 24 may transmit the slab to processing circuitry 22 for further processing. Note that in another embodiment, the processing circuitry may receive the slab from a data store.

[0022] The camera model defines the viewing direction. For example, in a simple camera model, a ray originates from the camera's center of projection and penetrates pixels on the image plane between the camera and the 3D volume. Any type of camera model can be used, such as a fisheye camera, a compound view camera, or a curved planar reconstruction (CPR)-based camera of other camera models. CPR-based camera models can include a projected CPR-based camera model, a stretched-CPR-based camera model, a straightened-CPR-based camera model, etc.

[0023] In this embodiment, processing circuitry 22 includes rendering circuitry 26 that projects the slab onto a two-dimensional (2D) image. For example, rendering circuitry 26 projects the slab onto the 2D image using one or more projection algorithms, such as the MIP algorithm, MinIP algorithm, AveIP algorithm, and / or other projection algorithms described above.

[0024] In this embodiment, processing circuitry 22 includes display circuitry 28 that causes 2D images to be displayed on display screen 18 to a user.

[0025] In this embodiment, the processing circuitry 22, image processing circuitry 24, rendering circuitry 26, and display circuitry 28 are implemented in a CPU and / or GPU, respectively, by a computer program containing computer-readable instructions that cause one or more operations of the medical imaging device 10 and / or the methods described in the embodiments to be performed. In another embodiment, these circuits may be implemented as one or more application specific integrated circuits (ASICs) or field programmable gate arrays (FPGAs).

[0026] Computing unit 12 also includes a hard drive and other components of a PC, such as RAM, ROM, a data bus, an operating system including various device drivers, and hardware devices including a graphics card, etc. For clarity, Figure 1 does not show these components.

[0027] FIG. 2A is a flowchart showing an overview of a process according to one embodiment.

[0028] In a first step 30, processing circuitry 22 receives a slab containing a number of samples specified by a camera model.

[0029] In step 32, processing circuitry 22 determines whether one or more first samples of the plurality of samples are part of an anatomical region of interest. An anatomical region of interest may also be referred to as an anatomical structure of interest. Examples of anatomical regions of interest include, but are not limited to, gaps or spaces within a tissue and / or between different tissues, cylindrical structures such as generally cylindrical structures, spherical structures such as generally spherical structures, or other structures. For example, gaps or spaces within a tissue include fractures or gaps or spaces between at least two bones. Cylindrical structures include blood vessels or portions of blood vessels. Spherical structures include abnormalities, etc.

[0030] In steps 34a and 34b, processing circuitry 22 projects the slab along the line of sight onto an image plane to form an image, also referred to as a two-dimensional (2D) image.

[0031] If it is determined that the first sample is not part of the anatomical region of interest, processing circuitry 22 projects multiple samples in step 34a using a projection algorithm, such as a MIP algorithm, a MinIP algorithm, an AveIP algorithm, or other projection algorithm.

[0032] If processing circuitry 22 determines that the first sample is part of the anatomical region of interest, then processing circuitry 22 projects the first sample using a first projection mode at step 34b, and projects one or more second samples of the plurality of samples that are not part of the region of interest using a second projection mode.

[0033] In some embodiments, the first and second projection modes are the same. In other embodiments, the first and second projection modes are different. The first projection mode includes at least one of a MIP algorithm, a MinIP algorithm, an AveIP algorithm, or another projection algorithm. The second projection mode includes at least one of a MIP algorithm, a MinIP algorithm, an AveIP algorithm, or another projection algorithm.

[0034] In embodiments where the first and second projection modes are different, at least one of the first and second projection modes comprises a MinIP algorithm and at least the other comprises a MIP algorithm.

[0035] In embodiments where the first and second projection modes are the same, processing circuitry 22 projects the first sample separately from the second sample. For example, processing circuitry 22 projects the first sample using the AveIP algorithm, which includes only averaging the voxel data corresponding to the first sample. Processing circuitry 22 also projects the second sample using the AveIP algorithm, which includes only averaging the voxel data corresponding to the second sample.

[0036] In step 36, processing circuitry 22 displays the image. For example, display circuitry 28 displays the image on display screen 18, although in some embodiments, the image may not be displayed. For example, in such embodiments, the image may be further processed, stored, and / or transmitted to other computing devices.

[0037] FIG. 2B illustrates other steps that may comprise the process shown in FIG. 2A.

[0038] In step 32a, processing circuitry 22 defines a threshold, range, or value of interest for a first measurement of the anatomical region of interest, for example, in response to a user instruction.

[0039] In embodiments, the target value of the first measurement value comprises the maximum value of the first measurement value, but in other embodiments, the target value may comprise a different value, such as the average value of the first measurement value or another value.

[0040] In instances where the anatomical region of interest includes a gap or space, the first measurement may include a dimension or size of the gap or space, and a threshold, range, or target value identifies which first sample of which dimension or size of gap or space is projected using the first projection mode.

[0041] In the example where the anatomical region of interest includes a cylindrical structure, the first measure may include vesselness, vessel bifurcation, or other measure.

[0042] In instances where the anatomical region of interest includes a spherical structure, the first measure may include a texture measure or other measure, such as sphericity and / or variance. The first measure may also be referred to as a metric. "Sphericity" may also be considered a measure of how spherical the anatomical region of interest is.

[0043] The threshold, range, or target value of the first measurement can be defined based on a selected window width and / or window level applied to the 2D image. For example, the window width encompasses a selected range of voxel values ​​of the slab. The window level can be considered the midpoint of a range of voxel values, e.g., the window width. The lower limit of the window width corresponds to the window level minus half the window width, and the upper limit of the window width corresponds to the window level plus half the window width. For example, the threshold or range can be selected to fall between 25% and 50% of the selected range of voxel values. The window level can also be referred to as W / L.

[0044] The selected range of voxel values ​​is related to the tissue type of the anatomical region of interest, including soft tissues such as muscle, tendons, ligaments, fat, fibrous tissue, blood vessels, and other soft tissues, and hard tissues such as bone and other hard tissues. For an example region of interest that includes hard tissues such as bone, the selected range of voxel values ​​includes bone voxel values. However, the present disclosure is not limited to anatomical regions of interest that include hard tissues such as bone.

[0045] In step 32b, processing circuitry 22 detects candidate anatomical regions, which are detected anatomical regions or structures that may be evaluated to determine whether they are anatomical regions of interest.

[0046] FIG. 3 illustrates an example of a candidate anatomical region 33. The candidate anatomical region 33 includes the first sample described above. The first sample is indicated by the reference numeral 33a in FIG. 3. In the embodiment of FIG. 3, the candidate anatomical region 33 includes twelve first samples 33a, but in other embodiments, the candidate anatomical region may include more or fewer than twelve first samples 33a. Processing circuitry 22 draws multiple lines in multiple directions from at least one first sample 33a. The multiple lines are indicated by arrows in FIG. 3. The multiple directions may include multiple random directions, a random uniform distribution of directions on a sphere, an intentional uniform distribution of directions on a sphere, such as the Fibonacci sphere illustrated in FIG. 3, or other distributions of directions. The multiple directions are different from the viewing direction defined by the camera model. Although the multiple lines are drawn from a single first sample 33a in FIG. 3, processing circuitry 22 may also draw multiple lines in multiple directions from individual first samples 33a. The first sample 33a is associated with a first type of tissue or a first fluid. The processing circuit 22 can detect one or more other first samples 33a associated with the same first type of tissue or fluid. The processing circuit 22 can also detect one or more second samples 33b associated with a second type of tissue or a second fluid adjacent to the first type of tissue or the first fluid. The second type of tissue or the second fluid is different from the first type of tissue or the first fluid. The detected second samples 33b indicate a candidate anatomical region 33. In examples where the candidate anatomical region (and the anatomical region of interest) includes a gap or space, the gap or space is filled with the first tissue or the first fluid. For example, a crack in a bone may be filled with fluid. That is, the first fluid includes the fluid in a fracture, and the second tissue includes the bone. One or each first sample 33a of the candidate anatomical region 33 may be referred to as a potential gap candidate.

[0047] In an example where the anatomical region of interest includes a cylindrical structure, processing circuitry 22 detects candidate anatomical regions using a vessel likelihood filtering method. A vessel likelihood filtering method that can be used by processing circuitry 22 is described, for example, in A.F. Frangi et al. (1998) "Multiscale vessel enhancement filtering," Medical Image Computing and Computer-Assisted Intervention - MICCAI'98, Lecture Notes in Computer Science, vol. 1496 - Springer Verlag, Berlin, Germany, pp. 130-137. The vessel likelihood filtering method can obtain a vessel likelihood measure or vessel branching based on all eigenvalues ​​of a local Hessian matrix. For example, the vessel likelihood filtering method uses a Gaussian derivative kernel to form a scale-specific Hessian matrix. Processing circuitry 22 can identify vessel likelihood or vessel branching by evaluating the eigenvalues ​​of the Hessian matrix. The obtained vessel likelihood or vessel branching indicates a candidate anatomical region.

[0048] In examples where the anatomical region of interest includes a spherical structure, the candidate anatomical region is detected based on a sphericity measurement or the like. For example, processing circuitry 22 may use or perform a spherical filtering method to detect candidate anatomical region 33. The spherical filtering method includes generating, for each first sample 33a of the candidate anatomical region 33, a uniform set of points on a spherical surface 34 centered on the first sample 33a. FIG. 4A is a schematic diagram illustrating the generated spherical surface, represented by a circle. In FIG. 4A, the center of the spherical surface 34 is indicated by the symbol P1. For clarity, FIG. 4A shows only one first sample 33a, but the candidate anatomical region 33 may include two or more first samples 33a and / or two or more generated spherical surfaces. For example, processing circuitry 22 may generate multiple overlapping spherical surfaces, which define a dense arrangement or grid of spherical surfaces.

[0049] In the embodiment of FIG. 4A , the set of points consists of eleven points P2, although in other embodiments, the set of points may include more or fewer than eleven. The sphere 34 has a constant radius R. The radius R depends on the expected size or dimension of the anatomical region of interest. The spherical filtering method further includes identifying or generating a gradient vector 35 for each point P2 on the sphere 34. Each identified gradient vector 35 indicates the direction or degree of change in voxel values. Each identified gradient vector 35 is identified or generated at each point P2 by a method or process that identifies or generates gradients at one or more subvoxel locations. A subvoxel location is a location between at least two voxels. For example, in an embodiment, each identified gradient vector 35 is generated at each point P2 based on a center difference approximation using an interpolation function, such as a trilinear interpolation function. Volume data between at least two voxels is interpolated using the interpolation function.

[0050] In other embodiments, each gradient vector is generated using a gradient reconstruction method, such as a direct gradient reconstruction method, which uses one or more partial derivatives of one or more quadratic or cubic interpolating polynomials, such as b-splines, Catmull-Rom splines, or other spline functions.

[0051] In an embodiment, processing circuitry 22 detects candidate anatomical regions based on the orientation of each identified gradient vector 35 of each point P2 relative to the spherical surface 34. For example, an orientation of each identified gradient vector 35 of each point P2 toward or away from the center P1 of the spherical surface 34, i.e., along the radius R of the spherical surface 34, indicates a candidate anatomical region having, for example, a spherical structure. In such an embodiment, the identified gradient vectors 35 are considered to be arranged radially on the spherical surface 34. Deviation of one or more of the identified gradient vectors 35 from an orientation along the radius R of the spherical surface 34 may indicate a decrease in sphericity. In an embodiment, the spherical filtering method includes identifying or generating an expected gradient vector 36 for each point P2.

[0052] In an embodiment, processing circuitry 22 generates each predicted gradient vector 36 for each point P2 such that the predicted gradient vector 36 extends along a direction of radius R toward or away from the center P1 of the spherical surface 34. In such an embodiment, the predicted gradient vectors 36 are considered to be disposed radially about the spherical surface 34.

[0053] In another embodiment, processing circuitry 22 determines a normalization vector based on the difference between the voxel value of the center P1 of first sample 33a and the voxel value of point P2, and generates a predicted gradient vector 36 for each point P2 by applying the normalization vector to a defined spherical model.

[0054] FIG. 4B is a schematic diagram illustrating a portion of a generated spherical surface 34 containing multiple predicted gradient vectors 36. The spherical filtering method may include determining an angle α between each predicted gradient vector 36 and the identified gradient vector 35 of each point P2. Based on the determined angle α for each point P2 on the spherical surface 34, processing circuitry 22 determines whether the first sample is at the center of the spherical structure. For example, if the angle between each identified gradient vector 35 and each predicted gradient vector 36 is approximately zero, the first sample at the center of the spherical surface 34 is the center of the spherical structure. One or more first samples that make up or are within the spherical surface 34 are considered to be part of the spherical structure, i.e., they may indicate a potential anatomical region. Additionally, the angle α for each point P2 on the spherical surface indicates the variance of the voxel value gradients of the spherical surface.

[0055] It should be noted that in other embodiments, the detection of candidate anatomical regions may occur separately from the process shown in FIG. 2B. In such embodiments, the candidate anatomical regions are identified by segmentation. For example, the processing circuitry or image processing circuitry detects or identifies the candidate anatomical regions according to a segmentation process or method.

[0056] In step 32c of this embodiment, processing circuitry 22 determines a second measurement of the candidate anatomical region.

[0057] For example, as described above with respect to the first measurement, in an example where the candidate anatomical region includes a gap or space, the second measurement may include a dimension or size of the gap or space. For example, as described above with respect to the first measurement, in an example where the candidate anatomical region includes a cylindrical structure, the second measurement may include vascularity, vascular branching, or other measurement. For example, as described above with respect to the first measurement, in an example where the candidate anatomical region includes a spherical structure, the second measurement may include a texture measurement, such as sphericity and / or variance, or other measurement. Note that the anatomical region of interest and the candidate anatomical region include the same type of anatomical region or structure. For example, both the anatomical region of interest and the candidate anatomical region include a gap or space, a cylindrical structure, a spherical structure, or other anatomical structure or region, as described above.

[0058] In an example where the candidate anatomical region includes the gap or space described above, a distance L between the first samples, e.g., at least two second samples, indicates the size or dimension of the gap or space. Distance L is shown in FIG. 3. For example, processing circuitry 22 may draw multiple lines from each first sample 33a, as described above, to identify the size or dimension of the gap or space.

[0059] In some embodiments, processing circuitry 22 determines the size or dimension of a gap or space based on a range of first measurements of the anatomical region of interest. This process may include measuring the position of the nearest first sample above the lower limit of the range and the position of the nearest first sample below the upper limit of the range and determining the minimum distance between the first samples. The minimum distance corresponds to or represents the minimum size or dimension of the gap or space. However, in other embodiments, processing circuitry 22 may determine, for example, an average size or dimension or a maximum size or dimension of the gap or space based on all first samples within the gap or space. Note that processing circuitry is not limited to determining the size or dimension of the gap or space using the processes described herein. For example, in other embodiments, processing circuitry may determine the size or dimension of the gap or space using other processes or methods.

[0060] In examples where the candidate anatomical region includes a cylindrical structure, processing circuitry 22 may use the vessel likelihood filtering method described above to determine the vessel likelihood and vascular branching of the candidate anatomical region. For example, processing circuitry 22 may use the vessel likelihood filtering method to determine the degree to which the candidate anatomical region is cylindrical.

[0061] In examples where the candidate anatomical region includes spherical structures, processing circuitry 22 determines the sphericity of the candidate anatomical region using the spherical filtering method described above. For example, processing circuitry 22 determines the sphericity of the candidate anatomical region based on the determined angle between each predicted gradient vector 36 and each gradient vector 35 at each point P2.

[0062] In other embodiments, determining the second measurement value of the candidate anatomical region may be performed separately from the process shown in FIG. 2B . In such embodiments, the processing circuitry may receive volume data including the candidate anatomical region. The volume data may also include a predetermined second measurement value of the candidate anatomical region. The volume data including the candidate anatomical region and the second measurement value of the candidate anatomical region may be acquired separately, for example, from a previous medical scan. For example, the processing circuitry may receive the volume data from a data storage unit. The volume data may also be referred to as a scalar volume.

[0063] In step 32d, processing circuitry 22 determines whether the second measurement is less than, within a range of, or corresponds, eg, substantially corresponds to, a value of interest of the first measurement.

[0064] If, in step 32e, processing circuitry 22 determines that the second measurement value is less than, within, or corresponds to, e.g., substantially corresponds to, a threshold value of the first measurement value, processing circuitry 22 determines that the first sample of the candidate anatomical region is part of an anatomical region of interest. For example, processing circuitry 22 qualifies the candidate anatomical region as an anatomical region of interest.

[0065] The processing procedures described above with respect to steps 30 through 36 of FIGS. 2A and 2B may be performed by rendering circuitry 26.

[0066] In the processing procedure described above with reference to FIG. 2B , processing circuitry 22 determines whether one or more first samples are part of an anatomical region of interest based on a range, threshold, or target value of the first measurement. However, in other embodiments, processing circuitry may determine whether one or more first samples are part of an anatomical region of interest based on a segmentation mask of the anatomical region of interest or a portion thereof. For example, processing circuitry 22 may generate a segmentation mask of at least a portion or the entire anatomical region of interest using a machine learning model, such as a deep learning model. However, in other embodiments, processing circuitry may receive the segmentation mask from data storage 16. The segmentation mask can be generated using, for example, a deep learning segmentation model described in J. Wasserthal et al., (2023) “TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images,” Radiology: Artificial Intelligence - arXiv: 2208.05868v2.

[0067] FIG. 5A shows a 2D image of an anatomical region of interest. The 2D image in FIG. 5A is generated by projecting an MPR slab using an MIP algorithm. In the example shown in FIG. 5A, a bone 40 in the anatomical region of interest has a crack 40a. In the image of FIG. 5A, the bone outline 42 is clearly visible, but the crack 40a is barely visible, which can make detecting the crack 40a difficult.

[0068] FIG. 5B shows another 2D image of the anatomical region of interest. The 2D image in FIG. 5B was generated by projecting an MPR slab using the MinIP algorithm. The image shown in FIG. 5B and the image in FIG. 5A are based on the same MPR slab. In the image in FIG. 5B, the crack 40a is clearly visible, but the bone contour 42 is obscured.

[0069] FIG. 5C shows a 2D image of an anatomical region of interest generated by the processing described above with respect to FIGS. 2A and 2B. In this example, the first projection mode is different from the second projection mode. The first projection mode includes a MinIP algorithm, and the second projection mode includes a MIP algorithm. For example, a first sample is part of the fissure 40a and is projected using the MinIP algorithm. A second sample, which is not part of the fissure 40a, is projected using the MIP algorithm. In other words, the projection mode of the first sample is switched from the MIP algorithm to the MinIP algorithm. This allows the image of the fissure 40a to be improved while preserving the bone contour 42.

[0070] That is, by projecting a first sample in a first projection mode and a second sample in a second projection mode, improved detection and / or visualization of an anatomical region of interest can be achieved. In this example, the processes described herein can improve visualization of small or narrow anatomical structures or regions, such as narrow crevices. For example, the processes described herein can improve visualization of small anatomical structures or regions, such as crevices or other anatomical structures or regions, having dimensions less than 5 mm, such as between about 0.5 mm and about 2 mm. Thus, in this embodiment, the threshold for the first measurement is set to about 5 mm, and / or the range for the first measurement is set to between about 0.5 mm and about 2 mm. Note that the threshold and / or range for the first measurement are not limited to the exemplary values ​​disclosed herein. Improved visualization of small or narrow anatomical structures or regions can enable more accurate quantitative measurement of the anatomical structures or regions. For example, improved visualization of small or narrow anatomical structures or regions can more accurately identify the extent and / or width of a crevices.

[0071] As can be seen in Figure 5A, the gaps and spaces between different bones can also be more clearly visualized. An example of such a gap or space is shown at 40b. Note that the above process can be performed for multiple anatomical regions of interest that are part of or contained within the slab, but for clarity, only fissure 40a will be discussed below.

[0072] 5A-5C, the pixels corresponding to the crack 40a have a lower intensity than the pixels corresponding to the bone 40. This is because the voxel values ​​corresponding to the crack 40a, e.g., the fluid within the crack (e.g., the first sample), are lower than the voxel values ​​corresponding to the bone 40 (e.g., the second sample).

[0073] However, in other embodiments, the intensity of one or more pixels corresponding to an anatomical region of interest may be higher than one or more pixels located outside the anatomical region of interest. For example, the anatomical region of interest may include an anomaly in the lungs of a patient or other subject. The anomaly may be shaped like a pocket and filled with fluid, such as air. In such embodiments, one or more voxel values ​​corresponding to the anatomical region of interest, such as a first sample, may be higher than one or more voxel values ​​corresponding to one or more second samples that are not part of the anatomical region of interest. Pixels with higher intensities are more likely to be visible as bright features or structures, also referred to as bridges, in the image.

[0074] 6A shows a 2D image of an anatomical region of interest generated by projecting an MPR slab using the AveIP algorithm. In the example of FIG. 6A, the anatomical region of interest includes the aforementioned cleft 40a. In this example, the cleft 40a is visible, but the extent and / or width of the cleft 40a is difficult to determine, and the outline 42 is blurred.

[0075] FIG. 6B illustrates another 2D image of an anatomical region of interest generated by the processing described above in connection with FIGS. 2A and 2B. In the example of FIG. 6B, the anatomical region of interest also includes the fissure 40a described above. In this example, the first projection mode is different from the second projection mode. The first projection mode includes a MinIP algorithm, and the second projection mode includes a MIP algorithm. For example, a first sample that constitutes the fissure 40a is projected using the MinIP algorithm, and a second sample that is not part of the fissure 40a is projected using the MIP algorithm.

[0076] In the example shown in FIG. 6B, the size or dimensions of the crack 40a are more accurately identified compared to those shown in FIGS. 5A-5C. For example, the threshold and / or range of the first measurement value used in the process of projecting the image of FIG. 6B is different, e.g., larger, than the threshold and / or range of the first measurement value used in the process of projecting the image of FIG. 5C. That is, more first samples are projected using the first projection mode compared to the example described with reference to FIG. 5C. Also, the width of the crack 40a is more accurately identified from the 2D image of FIG. 6B compared to the 2D image of FIG. 6A. Furthermore, the contour 42 in the 2D image of FIG. 6B is more clearly defined.

[0077] 7 shows a 2D image of a mask 44 defined by pixels corresponding to an anatomical region of interest. For example, processing circuitry 22 generates one or more first landmarks. The first landmarks indicate one or more pixels corresponding to a first sample projected in a first projection mode. Processing circuitry 22 generates one or more second landmarks. The second landmarks indicate one or more pixels corresponding to a second sample projected in a second projection mode. Processing circuitry 22 may also generate a mask based on the first and / or second landmarks. Mask 44 may be used in masking other images or overlaid on other images.

[0078] 7, processing circuitry 22 generates a mask 44 based on a first landmark. That is, mask 44 is defined only by pixels projected using a first projection mode, which in this example includes the MinIP algorithm. Mask 44 may be presented to a user as a tint of the intensity of pixels corresponding to the first and / or second samples.

[0079] 8 shows a 2D image of an anatomical region of interest generated by the processing described above with reference to FIGS. 2A and 2B. This anatomical region of interest also includes the fissure 40a. In this example, the first projection mode and the second projection mode are identical. The first and second projection modes include the AveIP algorithm. Processing circuitry 22 projects a first sample associated with fissure 40a separately from a second sample that is not part of fissure 40a.

[0080] To generate the 2D image of Figure 6A, the voxel data values ​​of all samples in the MPR slab were averaged. In contrast, in the example of Figure 8, the voxel values ​​of the first sample and the second sample that are part of the fissure 40a were averaged separately. In this manner, processing circuitry 22 may vary the samples used to project the image of fissure 40a.

[0081] 9 shows a 2D image of another anatomical region of interest generated by projecting an MPR slab using an MIP algorithm. The 2D image of FIG. 9 shows lungs 46 of a patient or other subject. In this embodiment, the region of interest includes an abnormality 48 in the lungs 46. In this embodiment, the abnormality 48 in the lungs 46 is due to emphysema. The abnormality 48 may have a spongy appearance. In this embodiment, the measurements of the anatomical region of interest include texture measurements. The texture measurements may be based on or include sphericity and / or variance.

[0082] FIG. 10 shows a 2D image of the anatomical region of interest of FIG. 9 , generated by the processing described above in connection with FIGS. 2A and 2B . In this embodiment, processing circuitry 22 defines a target value for the first measurement. In this embodiment, the target value is a sphericity value. In this embodiment, the target sphericity value is the maximum sphericity value of the anomalies 48. Processing circuitry 22 detects the anomalies 48 using the spherical filtering method described above. Processing circuitry 22 determines the sphericity of each anomaly 48 using the spherical filtering method described above. The determined sphericity is considered a second measurement. Processing circuitry 22 determines whether the determined sphericity of each anomaly 48 corresponds, e.g., substantially corresponds, to the target sphericity value. If the determined sphericity of each anomaly 48 corresponds, e.g., substantially corresponds, to the target sphericity value, processing circuitry 22 determines that the one or more first samples corresponding to the anomalies 48 are part of the anatomical region of interest.

[0083] As described above, processing circuitry 22 projects a first sample corresponding to anomaly 48 in a first projection mode. Processing circuitry 22 also projects one or more second samples that are not part of the region of interest in a second projection mode. In this embodiment, the first projection mode includes a MIP algorithm, and the second projection mode also includes a MIP algorithm. As described above with respect to FIG. 8 , if the first and second projection modes are the same, processing circuitry 22 projects one or more first samples and one or more second samples separately.

[0084] Based on the sphericity value of the object, processing circuitry 22 adjusts the window width and / or window level to select one of the anomalies 48 in the 2D image shown in FIG. 10. The anomaly 48 in the 2D image of FIG. 10 includes multiple spherical shapes. The region in FIG. 10 that includes the anomaly may be referred to as a grainy region. A large number of spherical shapes can be identified from the 2D image shown in FIG. 10, thereby enabling a quantitative measurement of the anomaly 48.

[0085] It should be noted that any of the above features are also applicable to embodiments in which the anatomical region of interest includes a cylindrical structure.

[0086] According to an embodiment, a medical visualization method is provided that includes a slab of samples identified by a camera / transform and a metric for detecting whether a sample is part of a gap. The method includes determining the length of the identified gap and two opposing intensity modes, such as MaxIP and MinIP, to operate on the samples of the slab. Each sample is considered a potential gap candidate. For a gap candidate, a gap length is identified, and if the length is within a range set by the user, the gap candidate is qualified as a gap. If there is a qualified gap sample, the intensity mode is switched from the default (MaxIP) to gap mode (MinIP), and non-gap samples are discarded.

[0087] It may detect bridges (high intensity close to gaps) and switch modes between different intensity modes.

[0088] During imaging, detection may be performed within the lines by drawing lines in different directions from the sample and measuring the gap length in that direction.

[0089] Gaps may be detected as segmented objects and the render pass may simply perform distance measurements.

[0090] Gap detection and determination of the minimum gap length for each voxel may be performed in advance and sent forward as a separate scalar volume.

[0091] All samples may be used for intensity projection after gap detection.

[0092] The pixels after the change may be written to a mask and used for overlay or other processing.

[0093] The two intensity modes may be identical, for example AvIP, except for the non-gap / bridge samples which are excluded.

[0094] The gap threshold may be defined relative to the W / L setting for the line of sight.

[0095] Gap detection may be modified to other geometric local measurements such as vessel-likeness or vessel branching, or texture metrics using image analysis-based segmentation mask coverage instead of intensity-based gap detection, such as sphericity, variance, and enhanced regions.

[0096] Although particular circuits are described herein, in other embodiments, the functionality of one or more of these circuits may be implemented by a single processing resource or other component. Alternatively, the functionality implemented by a single circuit may be implemented by a combination of two or more processing resources or other components. A single circuit includes multiple components, whether separate from one another, that implement the functionality of the circuit. Multiple circuits includes a single component that implements the functionality of those circuits.

[0097] Although several embodiments have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. The novel method and system may be embodied in various other forms, and various omissions, substitutions, and modifications may be made without departing from the spirit of the invention. These embodiments and modifications are intended to be included within the scope and spirit of the invention, as well as within the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0098] 10 Medical image processing device 12 Arithmetic unit 14 Scanner 16 Data storage unit 18 Display screen 22 Processing circuit 24 Image processing circuit 26 Rendering Circuit 28 Display circuit 30 Data storage unit

Claims

1. receiving a slab including a plurality of samples identified by the camera model; determining whether one or more first samples of the plurality of samples form part of an anatomical region of interest; forming an image by projecting the slab onto an image plane along a line of sight; If the processing circuitry determines that the one or more first samples form part of the anatomical region of interest, projecting the one or more first samples in a first projection mode; projecting one or more second samples of the plurality of samples that are not part of the anatomical region of interest in a second projection mode; Medical imaging equipment.

2. The first projection mode and the second projection mode are the same or different. The medical image processing device according to claim 1 .

3. the first projection mode includes at least one of a minimum intensity projection (MinIP) algorithm, a maximum intensity projection (MIP) algorithm, and an average intensity projection (AveIP) algorithm; the second projection mode includes at least one of a minimum intensity projection (MinIP) algorithm, a maximum intensity projection (MIP) algorithm, and an average intensity projection (AveIP) algorithm; The medical image processing device according to claim 1 .

4. If the first projection mode and the second projection mode are different, at least one of the first projection mode and the second projection mode includes a minimum intensity projection (MIP) algorithm, and at least the other includes a maximum intensity projection (MIP) algorithm. The medical image processing device according to claim 2 .

5. the processing circuit projects the one or more first samples separately from the one or more second samples when the first projection mode and the second projection mode are the same. The medical image processing device according to claim 2 .

6. If the first projection mode and the second projection mode are the same, the first projection mode and the second projection mode each include an average projection method (AveIP) algorithm. The medical image processing device according to claim 2 .

7. voxel values ​​corresponding to the one or more first samples are higher than voxel values ​​corresponding to the one or more second samples; or voxel values ​​corresponding to the one or more first samples are lower than voxel values ​​corresponding to the one or more second samples; The medical image processing device according to claim 1 .

8. The processing circuitry a threshold value of a first measurement of the anatomical region of interest; a range of the first measurement value of the anatomical region of interest; a target value of the first measurement value of the anatomical region of interest; specifying at least one of: The medical image processing device according to claim 1 .

9. the threshold, range, or target value of the first measurement is defined based on a selected window width and / or window level applied to the image; The medical image processing device according to claim 8 .

10. The first measurement of the anatomical region of interest comprises: the size or dimension of the anatomical region of interest; a vascularity of the anatomical region of interest or a vascular bifurcation of the anatomical region of interest; texture measurements; and including at least one of The medical image processing device according to claim 8 .

11. The processing circuitry determining a second measurement value of a candidate anatomical region including the one or more first samples or receiving a predetermined second measurement value of the candidate anatomical region; determining whether the second measurement is less than a threshold value of the first measurement, within a range of the first measurement, or corresponds to a target value of the first measurement; Execute at least one of the following: The medical image processing device according to claim 8 .

12. the processing circuit determines that the one or more first samples of the candidate anatomical region form part of the anatomical region of interest if the second measurement value is less than a threshold value of the first measurement value, is within a range of the first measurement value, or corresponds to a target value of the first measurement value. The medical image processing device according to claim 11 .

13. the processing circuitry detects the candidate anatomical regions. The medical image processing device according to claim 11 .

14. the processing circuitry detects the candidate anatomical region by drawing a plurality of lines in a plurality of directions from at least one of the one or more first samples and determines the second measurement of the candidate anatomical region. The medical image processing device according to claim 13 .

15. the processing circuitry determines whether the one or more first samples of the plurality of samples form part of the anatomical region of interest based on a segmentation mask of at least a portion of the anatomical region of interest. The medical image processing device according to claim 1 .

16. the candidate anatomical regions are identified by segmentation; The medical image processing device according to claim 11 .

17. the processing circuitry receives volume data including the candidate anatomical region and a predetermined second measure of the candidate anatomical region; The medical image processing device according to claim 11 .

18. one or more pixels of the image correspond to the one or more first samples; the one or more pixels define a mask that is used for masking or overlaying other images; The medical image processing device according to claim 1 .

19. the anatomical region of interest comprises a space or gap; the first measurement comprises a size or dimension of the space or gap; The medical image processing device according to claim 8 .

20. receiving a slab including a plurality of samples identified by a camera model; determining whether one or more first samples of the plurality of samples form part of an anatomical region of interest; and 1. A method of medical imaging, comprising forming an image by projecting the slab onto an image plane along a line of sight, The medical image processing method comprises: If it is determined that the one or more first samples form part of the anatomical region of interest, projecting the one or more first samples in a first projection mode; and projecting one or more second samples of the plurality of samples that are not part of the anatomical region of interest in a second projection mode. Medical image processing methods.

Images