Method, device, equipment, product and storage medium for identifying skull defect by constructing characteristic spectrum through nuclear magnetic image

Abstract

Belongs to the technical field of biomedical engineering, and discloses a method, a device, equipment, a product and a storage medium for identifying skull defects by constructing a characteristic spectrum through a nuclear magnetic image to solve the problem that a defect area is mistakenly divided into skull areas in existing automatic skull segmentation. Converting the human brain nuclear magnetic image to an anatomical space based on the plurality of anatomical points, and implementing preliminary segmentation to obtain a primary segmentation region of the skull; calculating coordinates of voxel points in the primary segmentation region in an anatomical point coordinate system, establishing a matrix, calculating a plurality of nuclear magnetic gray statistical characteristics of the primary segmentation region, and drawing a characteristic spectrum; and performing multi-condition feature threshold judgment on the feature map to obtain a logic binary map, and extracting a region of which a connected domain is greater than a specified threshold as a skull defect region. According to the method provided by the invention, a plurality of gray features are utilized, the misrecognition of the skull in the prior art is effectively corrected, the method can be further used for judging the filling tissue of the skull defect area, and the skull defect recognition capability can be improved.

Description

Technical Field

[0001] This application belongs to the field of biomedical engineering technology, specifically relating to methods, devices, equipment, products, and storage media for constructing feature maps from nuclear magnetic resonance images to identify skull defects. Background Technology

[0002] In medical image processing and clinical diagnosis, automatic segmentation of the human brain and skull from MRI images is a crucial foundation for 3D reconstruction, finite element simulation, and defect repair. With the development of imaging and artificial intelligence, existing automatic brain segmentation methods, primarily template-based registration and deep learning-based segmentation, have limitations in processing images of skull defect cases. This is because there is a lack of standardized atlases reflecting the characteristics of skull defects for registration reference, and a lack of optimized segmentation model construction strategies specifically tailored to skull defect conditions. The performance of supervised or semi-supervised networks is largely dependent on their training dataset, while skull defect datasets rarely contain a substantial number of samples. These limitations severely restrict the automatic identification and quantitative analysis of skull defects, making it difficult to meet the demands of accurate 3D reconstruction and finite element simulation.

[0003] To address the aforementioned issues, this application proposes a method for constructing a skull feature atlas using MRI images to effectively reflect the image feature representation of skull defects, and for identifying skull defects through this feature atlas. Summary of the Invention

[0004] To address the problems existing in the prior art, the purpose of this application is to provide a method for constructing a skull feature atlas based on MRI images to identify skull defects. The method involves converting human brain MRI images into images of anatomical space and performing preliminary screening of skull regions. The preliminary segmentation results are then mapped to the coordinate system of anatomical points for a second region screening. A matrix is ​​established in the coordinate system of anatomical points, and multiple MRI grayscale statistical features of the screened regions are calculated to construct a feature atlas. The feature atlas is then subjected to multi-condition feature threshold judgment and a second screening based on connected component thresholds to obtain accurate skull defect regions.

[0005] Specifically, this application relates to the following aspects: According to one aspect of this application, a method for constructing a skull feature atlas based on MRI images to identify skull defects is provided, comprising: converting a human brain MRI image to an anatomical space based on multiple anatomical points to obtain an anatomical space image; preliminarily segmenting the anatomical space image to obtain a primary segmentation region of the skull; establishing a matrix based on coordinates in the anatomical point coordinate system and calculating multiple MRI gray-level statistical features of the primary segmentation region therein, and drawing a feature atlas of the primary segmentation region using the multiple MRI gray-level statistical features to obtain multiple feature atlases; performing multi-condition feature threshold judgment on the multiple feature atlases to obtain a logical binary atlas; post-processing the logical binary atlases and extracting regions where connected components are greater than a specified threshold as skull defect regions, wherein the post-processing includes: closing operation, cavity filling, and connected component extraction.

[0006] According to some embodiments of this application, converting a human brain MRI image to anatomical space includes: spatially registering the human brain MRI image with a standard MRI template to obtain a linear registration matrix and a nonlinear registration matrix; using the linear registration matrix and the nonlinear registration matrix to calculate the world coordinates of multiple anatomical points in the standard MRI template in the human brain MRI image; and using the world coordinates of the multiple anatomical points to calculate an anatomical space transformation matrix to convert the human brain MRI image into an anatomical space image.

[0007] According to some embodiments of this application, calculating the anatomical space transformation matrix using the world coordinates of multiple anatomical points includes: calculating a first direction transformation matrix using the world coordinates of multiple anatomical points, and calculating the first coordinates of the multiple anatomical points after undergoing a first direction transformation using the first direction transformation matrix; calculating a second direction transformation matrix using the first coordinates of the multiple anatomical points, and calculating the second coordinates of the multiple anatomical points after undergoing a first direction transformation and a second direction transformation sequentially using the second direction transformation matrix; calculating a third direction transformation matrix using the second coordinates of the multiple anatomical points, and calculating the third coordinates of the multiple anatomical points after undergoing a first direction transformation, a second direction transformation, and a third direction transformation sequentially using the third direction transformation matrix; and constructing the anatomical space transformation matrix using the first direction transformation matrix, the second direction transformation matrix, the third direction transformation matrix, and the third coordinates.

[0008] According to some embodiments of this application, calculating the coordinates of each voxel point in the primary segmentation region in the anatomical point coordinate system includes: establishing a spherical coordinate system with the previous joint point as the origin of the anatomical point coordinate system; and calculating the polar angle and azimuth angle of each voxel point in the primary segmentation region in the spherical coordinate system as the coordinates of each voxel point in the anatomical point coordinate system.

[0009] According to some embodiments of this application, establishing a matrix based on coordinates in the anatomical point coordinate system and calculating multiple NMR grayscale statistical features of the primary segmented regions therein includes: using the change in polar angle... and the change in azimuth A two-dimensional matrix is ​​established with a step size. Multiple NMR gray-level statistical features of the primary segmentation region corresponding to each point in the two-dimensional matrix are calculated. These multiple NMR gray-level statistical features include the proportion of voxels below the median gray value of the skull, the average gray value, the lower quartile of gray value, the gray value variance, the gray value coefficient of variation, and the distance between voxels and multiple anatomical points.

[0010] According to some embodiments of this application, drawing a feature map of a primary segmentation region using multiple NMR grayscale statistical features includes: drawing a feature map for each of the multiple NMR grayscale statistical features; and combining the feature maps of all NMR grayscale statistical features as the feature map of the primary segmentation region.

[0011] According to another aspect of this application, an apparatus for constructing a skull feature atlas based on MRI images to identify skull defects is provided, characterized by comprising: a preprocessing unit, which converts a human brain MRI image to an anatomical space based on multiple anatomical points to obtain an anatomical space image, and preliminarily segments the anatomical space image to obtain a primary segmentation region of the skull; an atlas construction unit, which calculates the coordinates of each voxel point in the primary segmentation region in the anatomical point spherical coordinate system, establishes a matrix based on the coordinates in the anatomical point coordinate system and calculates multiple MRI grayscale statistical features of the primary segmentation region therein, draws a feature atlas of the primary segmentation region using the multiple MRI grayscale statistical features to obtain multiple feature atlases, and performs multi-condition feature threshold judgment on the multiple feature atlases to obtain a logical binary atlas; and a second extraction unit, which postprocesses the logical binary atlases and extracts regions where the connected components are greater than a specified threshold as skull defect regions, wherein the postprocessing includes: closing operation, cavity filling, and connected component extraction.

[0012] According to another aspect of this application, an electronic device is provided, comprising: a processor; and a memory storing program instructions that, when executed by the processor, cause the processor to perform the aforementioned method of constructing a skull feature atlas based on MRI images to identify skull defects.

[0013] According to another aspect of this application, a computer program product is provided, which includes computer program instructions that, when executed by a processor, cause the processor to perform the aforementioned method of constructing a skull feature map based on MRI images to identify skull defects.

[0014] According to another aspect of this application, a computer-readable storage medium is provided that stores computer program instructions, which, when executed by a processor, cause the processor to perform the aforementioned method for constructing a skull feature atlas based on MRI images to identify skull defects.

[0015] The method, apparatus, equipment, product, and storage medium provided in this application receive MRI images of the human skull, project the preliminary three-dimensional skull segmentation results onto a two-dimensional plane, and then use grayscale statistical feature analysis to determine the skull defect area. This allows for rapid and accurate determination of the presence of skull defects within the initially segmented area, effectively overcoming the problem of insufficient skull recognition accuracy in patients with skull defects in existing automatic segmentation methods. Furthermore, through in-depth analysis of the grayscale features of the defect area, the method provided in this application can also be used to further determine the type of filling tissue within the defect area, providing richer information support for clinical diagnosis and repair design. Attached Figure Description

[0016] Figure 1 The illustration shows a flowchart of a method for constructing a skull feature map based on MRI images to identify skull defects according to an embodiment of this application.

[0017] Figure 2 The illustration shows a schematic diagram of a method for identifying skull defects by constructing a skull feature map based on MRI images, according to an embodiment of this application, to segment skull defects.

[0018] Figure 3 The illustration shows a block diagram of an apparatus for constructing a skull feature map based on MRI images to identify skull defects, according to an embodiment of this application.

[0019] Figure 4 A block diagram of an electronic device according to an embodiment of this application is illustrated. Detailed Implementation

[0020] The present application is further illustrated below with reference to embodiments. It should be understood that the embodiments are only used to further illustrate and explain the present application and are not intended to limit the present application.

[0021] Unless otherwise defined, technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art. While similar or identical methods and materials may be applied in experimental or practical applications, materials and methods are described herein. In case of conflict, the definitions included herein shall prevail. Furthermore, materials, methods, and examples are for illustrative purposes only and are not intended to be limiting. The present application is further described below with reference to specific embodiments, but is not intended to limit the scope of the application.

[0022] Application Overview As mentioned above, existing automatic segmentation methods for the human brain and skull mainly fall into two categories. The first category is registration methods based on segmentation templates. This method achieves automatic segmentation of the target region by spatially registering the image to be segmented with a standard segmentation template. However, existing segmentation templates are mostly constructed based on normal human brains (there are no widely applicable templates for skull defects). The skull structure in the templates is relatively complete. When processing images of patients with skull defects, template registration often misidentifies the defect area as normal skull, resulting in inaccurate segmentation of the defect area and affecting subsequent defect identification and repair design. The second category is segmentation methods based on deep learning. Deep convolutional neural network technology has been widely used in medical image segmentation. This technology learns image features through a training set and achieves automatic segmentation, such as Chinese patents CN107437251B and CN117372484A, which use neural networks to register and identify skull regions in images. However, most of the existing training sets are derived from normal brain images with intact skulls, and there is a lack of large-scale sample data for patients with skull defects. Therefore, the trained models often cannot accurately identify the defect area when dealing with cases of skull defects, and the segmentation performance is significantly reduced.

[0023] The purpose of the method, apparatus, equipment, product, and storage medium proposed in this application for constructing a skull feature atlas based on MRI images to identify skull defects is to address the problem of inaccurate skull segmentation results for skull defect samples in existing automatic segmentation processes of human brain MRI images due to registration errors or imbalances in positive and negative sample sizes. This method converts human brain MRI images to anatomical space based on multiple anatomical points. Through a three-dimensional mapping method, the coordinates of each voxel point in the anatomical coordinate system are calculated and projected onto a two-dimensional plane. In this two-dimensional plane, each two-dimensional region corresponds to a portion of the primary segmented region in the skull. By calculating the statistical characteristics of multiple grayscale values ​​in the MRI images of each primary segmented region, the presence of skull defects in that region is further determined, and the skull defect portion can be restored and output as a skull defect segmentation result. In this way, by quickly, effectively, and accurately acquiring the grayscale statistical characteristics of each primary segmented region, not only can misidentification of complete skull regions be corrected during automatic segmentation, but the filling tissue at the skull defect region can also be further determined based on the grayscale characteristics, thereby improving the accuracy of the final segmentation.

[0024] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0025] Exemplary methods Figure 1 The illustration shows a flowchart of a method for constructing a skull feature map based on MRI images to identify skull defects according to an embodiment of this application.

[0026] Reference Figure 1 The method for constructing a skull feature map based on MRI images to identify skull defects according to embodiments of this application includes the following steps.

[0027] Step S110 involves performing an anatomical space transformation on the human brain MRI image samples to obtain human brain MRI image samples in an anatomical space. The anatomical space is based on points and lines in one or more anatomical structures, aligning human brain MRI image samples from different populations into a standardized space where subtle differences in the size and shape of the human brain are normalized. In a preferred example provided in this application, the anatomical space is constructed based on two anatomical structures: the anterior commissure (AC) and the posterior commissure (PC). The space is constructed with either AC or PC as the origin of the coordinate system containing the anatomical space, and the line connecting the two as one of the coordinate axes of the coordinate system containing the anatomical space. Then, the coordinates of each point in the human brain MRI image in the anatomical space are calculated.

[0028] Specifically, this step includes steps S1101-S1103: Step S1101: Perform spatial linear and nonlinear registration between the human brain MRI image sample and the standard MRI template to obtain the linear registration matrix and the nonlinear registration matrix. The standard MRI template can be a recognized skull standard registration template such as the MNI152 template.

[0029] Step S1102: Using the obtained linear and nonlinear registration matrices, calculate the world coordinates of multiple landmark points in the anatomical space within the selected standard MRI template, as shown in the human brain MRI image sample. These multiple landmark points in the anatomical space are anatomical structural points that define the anatomical space coordinate system. In the preferred example given above, these landmark points preferably include the anterior commissure point in anatomy. A , post-combination point P and the dome D Specifically, the world coordinate system is a coordinate system constructed using one or more points in the real world. In this step, the former joint point is used as the origin of the world coordinate system, and its world coordinates on each coordinate axis are all 0. The world coordinates of the other two marker points depend on their distance from the former joint point in the human brain MRI image. The linear registration matrix and the nonlinear registration matrix contain the world coordinates of each point in the human brain MRI image sample in the world coordinate system of the standard registration template. By multiplying the anatomical space transformation matrix with the world coordinates of the marker points in the human brain MRI image in the standard registration template, the world coordinates of each point in the human brain MRI image sample in the world coordinate system established with the former joint point as the origin are calculated, thereby removing individual differences in the human brain MRI image samples to achieve normalization.

[0030] Step S1103: Calculate the anatomical space transformation matrix using the calculated world coordinates of the landmark points. affine And based on this matrix affine Transform human brain MRI images into anatomical space. In a preferred example, the anterior commissure point is selected. A , post-combination point P And the dome D Together they constitute the anatomical space ACPC via the pre-joint point A The transformation matrix is ​​calculated from the world coordinates obtained at the joint point P. middle, x Transformation matrix in direction The calculation method is as follows:

[0031] (Formula 1) in, , , ; and Representing points respectively A and points P z-axis coordinate, AP for A , P The distance between two points.

[0032] Then, through Calculate the marker points separately A , P , D go through x Marker points after axis transformation , , Based on , , calculate y Transformation matrix in direction The calculation method is as follows:

[0033] (Formula 2) in, , , ; and Representing points respectively and points of x Axis coordinates for x After axis transformation A , D The distance between two points.

[0034] Then, through Calculate the marker points separately , , go through y Marker points after axis transformation , , Based on , , Calculate the transformation matrix in the z-direction. The calculation method is as follows:

[0035] (Formula 3) in, , , ; and Representing points respectively and points of x Axis coordinates for y After axis transformation A , P The distance between two points.

[0036] Then, through Calculate the marker points separately , , The marker point after z-axis transformation .

[0037] Finally, an anatomical space transformation matrix is ​​constructed based on the above three transformation matrices. The calculation method is as follows:

[0038] (Formula 4) in, , The transformation matrix is ​​obtained. Then, matrix operations can be used to convert the data of human brain MRI image samples into anatomical space, forming anatomical space images.

[0039] Step S120: An automatic tissue segmentation tool is used to automatically segment the human brain MRI image sample in the anatomical space to obtain a primary segmentation region of the skull, which is then used as the skull segmentation result. The SSD automatic tissue segmentation tool is an existing image segmentation technology. It can use SPM, SIMNIBS, FreeSurfer, or other deep segmentation networks for MRI images. As mentioned earlier, due to the performance limitations of these existing tissue segmentation tools, the obtained primary segmentation region will contain some non-defective tissue, requiring further removal of these unnecessary tissues.

[0040] Step S130: Select one of the multiple landmarks in the anatomical space as the anatomical point, establish a coordinate system using this point as the origin, preferably a spherical coordinate system, and calculate the coordinate values ​​of each voxel in the primary segmentation region within this coordinate system. Following the preferred example above, select the anterior commissure point as the anatomical point to establish a spherical coordinate system, and calculate the coordinates of each voxel in the primary segmentation region within this spherical coordinate system. , ),in ( ) indicates polar angle, ( () indicates the azimuth angle.

[0041] Step S140, using coordinate change and A two-dimensional matrix is ​​established for the step size, mapping the primary segmentation region to a two-dimensional space, so that each point in the two-dimensional matrix contains a portion of the NMR grayscale statistical features of the primary segmentation region. Specifically, using... and After establishing a two-dimensional matrix for the step size, each point in the space containing the two-dimensional matrix corresponds to a set of multiple points in the anatomical space of a human brain MRI image sample within a certain range of polar and azimuth angles. The value of the point at that location represents the values ​​of multiple gray-level statistical information of the primary segmentation region composed of multiple points of the human brain MRI image sample within that certain range of polar and azimuth angles in the anatomical space. For example, and When both angles are 3°, the point [10. 20] in the two-dimensional matrix can be represented as... ∈(27, 30]、 The primary segmentation region consisting of the set of points within the range (57, 60) and its NMR grayscale statistical characteristics.

[0042] Step S150: Calculate the NMR grayscale statistical features within each primary segmentation region of the two-dimensional matrix. The NMR grayscale statistical features include (but are not limited to) the proportion of voxels below the skull grayscale median, the grayscale mean, the lower quartile of grayscale, the grayscale variance, the grayscale coefficient of variation, and / or the distance between points within the region and one or more marker points (e.g., the anterior joint point in the aforementioned preferred example). A feature map is then drawn based on each statistical feature.

[0043] Step S160 involves combining multiple feature maps of each primary segmentation region to perform multi-condition feature threshold judgment, thereby converting the multiple feature maps into a single logical binary map. The multi-condition feature threshold judgment refers to setting multiple statistical feature thresholds based on the multiple gray-level statistical features in step S150 to construct a rule set. Based on the rule set, feature maps containing specific gray-level statistical features (or combinations thereof) within the feature threshold range are retained, so that the ultimately retained feature maps collectively constitute a single binary map. In a preferred example, the rule set includes: below the 10th quantile of voxels with gray levels lower than the median gray level of the skull; below the 10th quantile of the average gray level; below the 10th quantile of the gray-level variance; below the 10th quantile of the difference between the upper and lower quartiles of gray level; and / or below the 10th quantile of the gray-level coefficient of variation. In other words, feature maps that satisfy the above five statistical values ​​(including the values ​​calculated by combining the statistical values) below the 10th percentile in multiple feature maps are retained to form a binary map, one of which contains information about the tissue of the skull defect.

[0044] In step S160, this application also provides another preferred example for filtering multiple feature maps as a supplement / alternative to the above preferred example. Specifically, the rule set includes: a voxel with a gray level lower than the median gray level of the skull, a voxel with a gray level lower than the median gray level, a voxel with a gray level lower than the median gray level, a voxel with a gray level lower than the median gray level, a voxel with a gray level lower than the median gray level, and / or a voxel with a gray level lower than the median gray level, and / or a voxel with a distance difference between the voxel and the anterior commissure point on the contralateral skull. That is, feature maps that satisfy the above five statistical values ​​(including the values ​​calculated by combining the statistical values) and are higher than the 15th percentile are retained to form a binary map, one of which does not contain information about the tissue of the skull defect.

[0045] Step S170 involves performing graphical processing on the obtained binary logistic map, including closing operations, hole filling, and connected component extraction, to filter out the skull defect segmentation results. Specifically, regions with connected components larger than a connected component threshold are extracted as the skull defect segmentation results. Based on the preferred example of filtering multiple feature maps in step S160 above, the connected component threshold can be 200, and the portion of the map with a connected component threshold greater than 200 after closing and hole filling is selected as the final skull defect segmentation result.

[0046] Finally, the method may further include step S180, which converts the skull defect segmentation result in the binary atlas into a skull segmentation mask on a three-dimensional image, as the final skull defect identification result that can be viewed by medical personnel or researchers. Step S180 can be performed using geometric mapping, projection mapping, or a neural network pre-trained on MR skull images to observe the skull of the sample and its defects.

[0047] The following reference Figure 2 This paper provides a preferred example of using the method described in this application to construct a skull feature atlas based on MRI images to identify skull defects in a given MRI image sample. (a) shows the acquired original human brain MR image sample. This sample is characterized by its physical space origin not being located at the geometric center of the brain, and the brain being asymmetrical about the origin, which can easily lead to significant calculation errors in subsequent calculations. Therefore, the origin coordinates of the sample's physical space are first adjusted to the center of the brain by setting the anatomical space: the anterior commissure point is selected as the coordinate origin, and the line connecting the anterior and posterior commissure points is used as the spatial origin. y (a) The original human brain MR image sample is adjusted into the ACPC anatomical space using the method provided in step S110 above, so as to eliminate the influence of individual physical structure differences of the sample. (b) shows the image after the original human brain MR image sample is adjusted into the ACPC anatomical space. In the anatomical space, the origin of the physical space coordinates of this sample is located at... of A The point is the center of the red cross line in (b).

[0048] Furthermore, (c) shows the... The primary segmented region of the skull obtained after SPM automatic segmentation, where the green area is the skull mask obtained from the initial segmentation. (d) Displays multiple feature maps of the primary segmented region in a two-dimensional matrix based on the change in voxel position coordinates in a spherical coordinate system with the previous joint point as the origin. These feature maps include the proportion of voxels below the median gray value of the skull, the average gray value, the lower quartile of gray value, the gray value variance, the gray value coefficient of variation, and the distance between points in the region and the previous joint point. The horizontal axis of these maps represents the azimuth angle of the voxel point with the previous joint point as the origin of the spherical coordinate system. The vertical axis represents the polar angle of the voxel point with the previous joint point as the origin of the spherical coordinate system. ,set up and The step size is Therefore, the resolution of each image is 120*60.

[0049] Therefore, each pixel in the graph corresponds to the following: and The primary segmentation region of the range, the pixel value represents the NMR grayscale statistical features of the skull region. In (d), from top to bottom, are the proportion of voxels below the median grayscale value of the skull, the lower quartile of grayscale, and the average grayscale value (this is just for demonstration purposes; the actual statistical features used are not just these three, but the six mentioned above).

[0050] (e) shows the binary image segmentation result obtained by combining feature thresholds of multiple feature maps and constructing a rule set in the manner described in step S160 above. The red area represents the desired segmentation result, indicating the mapping of the skull defect area in the binary image. (f) and (g) show the final segmentation result after remapping the skull defect area in the two-dimensional image to three-dimensional space. The green area represents the area of ​​normal skull obtained by automatic segmentation, while the red area represents the skull defect area that was incorrectly segmented into a normal area by automatic segmentation and was additionally identified after constructing multiple feature maps using grayscale statistical features. (f) shows the result as a horizontal slice, and (g) shows the result from the overall three-dimensional perspective of the original human brain MR image sample. This result is consistent with the visual diagnosis result of the MR image.

[0051] As can be seen, existing skull segmentation methods tend to miss some skull defect areas to a certain extent. The method described in this application, however, effectively identifies these skull defect areas by constructing a feature map, based on existing segmentation methods, thus ensuring high recall for automatic skull defect identification. This method is a good supplement to existing skull segmentation methods.

[0052] Exemplary device Figure 3The illustration shows an apparatus for constructing a feature map based on MRI images to identify skull defects according to an embodiment of this application.

[0053] Reference Figure 3 The apparatus 200 for identifying skull defects based on nuclear magnetic resonance images, according to an embodiment of this application, includes the following components.

[0054] The preprocessing unit 210 converts the human brain MRI image to the anatomical space based on multiple anatomical points to obtain the anatomical space image, and preliminarily segments the anatomical space image to obtain the primary segmentation region of the skull; The map construction unit 220 calculates the coordinates of each voxel point in the primary segmentation region in the anatomical point spherical coordinate system, establishes a matrix based on the coordinates in the anatomical point coordinate system and calculates multiple NMR grayscale statistical features of the primary segmentation region therein, draws the feature map of the primary segmentation region using the multiple NMR grayscale statistical features, obtains multiple feature maps, performs multi-condition feature threshold judgment on the multiple feature maps, and obtains a logical binary map. Extraction unit 230 post-processes the logical binary graph and extracts regions with connected components greater than a specified threshold as skull defect regions. The post-processing includes: closing operation, hole filling, and connected component extraction.

[0055] Here, those skilled in the art will understand that the specific functions and operations of each unit in the above-described device 200 for constructing feature maps based on MRI images to identify skull defects have been referenced above. Figures 1-2 The method for constructing feature maps based on MRI images to identify skull defects has been described in detail, and therefore, its repeated description will be omitted.

[0056] As described above, the apparatus 200 for identifying skull defects by constructing feature maps based on MRI images according to embodiments of this application can be implemented in various terminal devices, such as servers for storing primary segmented regions, multiple feature maps, and logical binary maps. In some examples, the apparatus 200 for identifying skull defects by constructing feature maps based on MRI images according to embodiments of this application can be integrated into the terminal device as a software module and / or a hardware module. For example, the apparatus 200 for identifying skull defects by constructing feature maps based on MRI images can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the apparatus 200 for identifying skull defects by constructing feature maps based on MRI images can also be one of many hardware modules of the terminal device.

[0057] Alternatively, in other examples, the device 200 for constructing feature maps based on MRI images to identify skull defects may be a separate device from the terminal device, and the device 200 may be connected to the terminal device via wired and / or wireless networks and transmit interactive information in accordance with an agreed data format.

[0058] Exemplary electronic devices Below, for reference Figure 4 This describes an electronic device according to embodiments of the present application.

[0059] Figure 4 A block diagram of an electronic device according to an embodiment of this application is illustrated.

[0060] The electronic device 10 includes one or more processors 11 and memory 12.

[0061] The processor 11 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0062] The memory 12 may include one or more program products, which may include various forms of readable storage media, such as one or more computer program products or embedded program products and corresponding computer-readable storage media or embedded readable storage media. The readable storage media may be, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions or embedded program products may be stored on the readable storage medium. The processor 11 may execute the program instructions to implement the methods for constructing feature maps based on MRI images to identify skull defects and / or other desired functions described in the various embodiments of this application above. The readable storage medium may also store various contents such as human brain MRI images, primary segmentation regions, multiple feature maps, logical binary maps, and / or skull defect regions.

[0063] In one example, the electronic device 10 may also include an input device 13 and an output device 14, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0064] The input device 13 may include, for example, a keyboard, a mouse, etc.

[0065] The output device 14 can output various information to the outside, including stored logical binary maps and / or skull defect areas, etc. The output device 14 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0066] Of course, for the sake of simplicity, Figure 4 Only some of the components of the electronic device 10 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 10 may include any other suitable components depending on the specific application.

[0067] Exemplary computer program products and computer-readable storage media In addition to the methods and devices described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps of the method for constructing feature maps based on MRI images to identify skull defects according to embodiments of this application, as described in the "Exemplary Methods" section of this specification.

[0068] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C, Python, or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0069] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the method for constructing feature maps based on MRI images to identify skull defects according to embodiments of this application, as described in the "Exemplary Methods" section of this specification.

[0070] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0071] By using the method, apparatus, equipment, product and storage medium provided in this application for constructing feature maps from MRI images to identify skull defects, the preliminary segmentation results of the three-dimensional skull are projected onto a two-dimensional plane. Combined with gray-scale statistical feature analysis, it is possible to quickly and accurately determine whether there are skull defect features in the area obtained from the preliminary segmentation, effectively overcoming the problem of insufficient identification accuracy of traditional segmentation methods in patients with defects.

[0072] Furthermore, through in-depth analysis of the grayscale characteristics of the defect area, the methods, apparatus, equipment, products, and storage media provided in this application can not only identify the defect location, but also further determine the type of filling tissue within the defect area, providing richer information support for clinical diagnosis and repair design.

[0073] Furthermore, in the feature maps calculated by the methods, apparatus, equipment, products, and storage media provided in this application, each location truly corresponds to the actual physiological location of the human skull. Technicians can directly make a rapid and accurate judgment on the location of skull defects based on these maps, thereby improving the efficiency of clinical diagnosis.

[0074] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0075] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0076] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0077] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0078] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for constructing a skull feature atlas based on MRI images to identify skull defects, characterized in that, include: Human brain MRI images are converted to anatomical space based on multiple anatomical points to obtain anatomical space images; The anatomical spatial image is initially segmented to obtain the primary segmentation region of the skull; Calculate the coordinates of each voxel point in the primary segmentation region in the anatomical point coordinate system, establish a matrix based on the coordinates in the anatomical point coordinate system and calculate multiple NMR grayscale statistical features of the primary segmentation region, and use the multiple NMR grayscale statistical features to draw the feature map of the primary segmentation region to obtain multiple feature maps. Multi-condition feature threshold judgment is performed on the multiple feature maps to obtain a logical binary map; The logical binary graph is post-processed, and regions with connected components greater than a specified threshold are extracted as skull defect regions. The post-processing includes: closing operation, hole filling, and connected component extraction.

2. The method for constructing a skull feature atlas based on MRI images to identify skull defects according to claim 1, characterized in that, Converting human brain MRI images to anatomical space includes: The human brain MRI image is spatially registered with a standard MRI template to obtain a linear registration matrix and a nonlinear registration matrix. The world coordinates of the multiple anatomical points in the standard MRI template in the human brain MRI image are then calculated using the linear and nonlinear registration matrices. The anatomical space transformation matrix is ​​calculated using the world coordinates of the multiple anatomical points to convert the human brain MRI image into the anatomical space image.

3. The method for constructing a skull feature atlas based on MRI images to identify skull defects according to claim 2, characterized in that, Calculating the anatomical space transformation matrix using the world coordinates of the multiple anatomical points includes: A first direction transformation matrix is ​​calculated using the world coordinates of the plurality of anatomical points, and the first coordinates of the plurality of anatomical points after the first direction transformation are calculated using the first direction transformation matrix. A second direction transformation matrix is ​​calculated using the first coordinates of the plurality of anatomical points, and the second coordinates of the plurality of anatomical points after sequentially undergoing the first and second direction transformations are calculated using the second direction transformation matrix. A third-direction transformation matrix is ​​calculated using the second coordinates of the plurality of anatomical points, and the third coordinates of the plurality of anatomical points after sequentially undergoing a first-direction transformation, a second-direction transformation, and a third-direction transformation are calculated using the third-direction transformation matrix. The anatomical space transformation matrix is ​​constructed using the first direction transformation matrix, the second direction transformation matrix, the third direction transformation matrix, and the third coordinate.

4. The method for constructing a skull feature atlas based on MRI images to identify skull defects according to claim 1, characterized in that, Calculating the coordinates of each voxel point in the anatomical point coordinate system within the primary segmentation region includes: A spherical coordinate system is established by taking the joint point as the origin of the coordinate system of the anatomical point; The polar angle and azimuth angle of each voxel point in the primary segmentation region in the spherical coordinate system are calculated as the coordinates of each voxel point in the anatomical point coordinate system.

5. The method for constructing a skull feature atlas based on MRI images to identify skull defects according to claim 4, characterized in that, A matrix is ​​established based on the coordinates in the anatomical point coordinate system, and multiple NMR grayscale statistical features of the primary segmented regions are calculated, including: The change in polar angle and the change in azimuth Establish a two-dimensional matrix with step size; Calculate multiple NMR grayscale statistical features of the primary segmented region corresponding to each point in the two-dimensional matrix. The multiple NMR grayscale statistical features include the proportion of voxels below the median grayscale value of the skull, the average grayscale value, the lower quartile of grayscale, the grayscale variance, the grayscale coefficient of variation, and the distance between voxels and multiple anatomical points.

6. The method for constructing a skull feature atlas based on MRI images to identify skull defects according to claim 4, characterized in that, The feature map of the primary segmentation region is drawn using the multiple NMR grayscale statistical features, including: Based on each of the multiple NMR grayscale statistical features, a feature map is drawn. The feature map that combines all NMR grayscale statistical features is used as the feature map of the primary segmentation region.

7. A device for constructing a skull feature atlas based on MRI images to identify skull defects, characterized in that, include: The preprocessing unit converts the human brain MRI image to anatomical space based on multiple anatomical points to obtain an anatomical space image, and preliminarily segments the anatomical space image to obtain the primary segmentation region of the skull; The map construction unit calculates the coordinates of each voxel point in the primary segmentation region in the anatomical point spherical coordinate system, establishes a matrix based on the coordinates in the anatomical point coordinate system and calculates multiple NMR grayscale statistical features of the primary segmentation region, draws the feature map of the primary segmentation region using the multiple NMR grayscale statistical features, obtains multiple feature maps, performs multi-condition feature threshold judgment on the multiple feature maps, and obtains a logical binary map. The extraction unit post-processes the logical binary graph and extracts regions with connected components greater than a specified threshold as skull defect regions. The post-processing includes: closing operation, hole filling, and connected component extraction.

8. An electronic device, characterized in that, include: processor; The memory stores program instructions that, when executed by the processor, cause the processor to perform the method described in claims 1-6 for constructing a skull feature map based on MRI images to identify skull defects.

9. A computer program product, characterized in that, It includes computer program instructions that, when executed by a processor, cause the processor to perform the method described in claims 1-6 for constructing a skull feature map based on MRI images to identify skull defects.

10. A computer-readable storage medium, characterized in that, The system stores computer program instructions that, when executed by a processor, cause the processor to perform the method described in claims 1-6 for constructing a skull feature map based on MRI images to identify skull defects.

Images