A method, device and storage medium for extracting brain symmetry plane

Abstract

The present application discloses a method, device and storage medium for extracting brain symmetry planes. The method comprises: inputting the cranial brain image data of the brain symmetry plane to be extracted into a trained brain partitioning network model to perform a brain partitioning operation to obtain a brain partitioning result; sampling spatial coordinate points in the left brain area and the right brain area respectively according to the brain partitioning result to obtain corresponding sampling point coordinates; and inputting the sampling point coordinates into the brain symmetry plane extraction model to perform a symmetry plane extraction operation to extract the brain symmetry plane. Using the scheme of the present application, the symmetry plane of the brain can be segmented quickly and accurately on normal and abnormal cranial brain images.

Description

Technical Field

[0001] The present application generally relates to the field of artificial intelligence technology. More specifically, the present application relates to a method, device and computer-readable storage medium for extracting brain symmetry planes. Background Art

[0002] With the development of computer technology, medical imaging is increasingly used in clinical practice. Medical imaging provides a great scientific and intuitive basis for the identification and treatment of diseases. It can better cooperate with clinical symptoms, tests, etc., and play an irreplaceable role in the final accurate identification and treatment of the disease. For cranial and brain diseases, since the brain is symmetrical, it can be used as a basis for judging the occurrence of cranial and brain lesions. For example, when a lesion occurs on one side of the brain, the scope and severity of the lesion can be determined by comparing the cranial and brain imaging data with the opposite side. Therefore, it is crucial to obtain the cranial and brain symmetry axis / symmetry plane of the cranial and brain imaging data.

[0003] At present, the extraction of cranial symmetry axis / symmetry plane of cranial brain image data is often carried out by manual drawing or template registration. Among them, the manual drawing method is subjective and depends on the experience of the observer, so there are large errors and low efficiency. The template registration method is to align the image to be processed with the standard brain template image, apply the registration matrix to the symmetry plane of the template, and obtain the symmetry plane of the image to be processed. Because this method is based on image registration technology, it has a long iteration time and slow processing speed. In addition, this method is highly dependent on the template. In actual clinical cases, many patients have asymmetric brain morphology or serious lesions (such as hematoma, infarction, skull defect, etc.), which leads to inaccurate registration results of this type of cranial brain images, resulting in poor extraction of symmetry planes and low algorithm stability.

[0004] In view of this, there is an urgent need to provide a solution for extracting brain symmetry planes so as to quickly and accurately segment the brain's symmetry planes on normal and abnormal cranial brain images. Summary of the invention

[0005] In order to at least solve one or more of the above-mentioned technical problems, the present application proposes a solution for extracting brain symmetry planes in multiple aspects.

[0006] In a first aspect, the present application provides a method for extracting brain symmetry planes, comprising: inputting cranial brain image data of the brain symmetry plane to be extracted into a trained brain partitioning network model to perform a brain partitioning operation to obtain a brain partitioning result; sampling spatial coordinate points in the left brain area and the right brain area according to the brain partitioning result to obtain corresponding sampling point coordinates; and inputting the sampling point coordinates into a brain symmetry plane extraction model to perform a symmetry plane extraction operation to extract the brain symmetry plane.

[0007] In one embodiment, the brain partition network model is trained by the following operations to obtain the trained brain partition network model: the cranial imaging data with labeled left and right brain regions are divided into a training set, a validation set, and a test set; the brain partition network model is trained using the training set to obtain a target partition result; and the brain partition network model is validated and tested using the validation set and the test set to obtain the trained brain partition network model.

[0008] In another embodiment, the trained brain partition network model includes an encoder and a decoder, and the encoder performs downsampling and outputs features related to brain partitions, and the decoder takes the features output by the encoder as input, performs upsampling and outputs the brain partition results.

[0009] In yet another embodiment, the brain image data at least includes brain CT image data and / or brain MR image data.

[0010] In another embodiment, the sampling point coordinates are input into a brain symmetry plane extraction model to perform a symmetry plane extraction operation to extract the brain symmetry plane, including: inputting the sampling point coordinates into a brain symmetry plane extraction model to perform a symmetry plane extraction operation to obtain a target weight vector, a target bias, and the coordinates of support vector points, wherein the support vector points are pairs of sampling points with the smallest distance between the midpoints of the left and right brain regions and the largest distance from the brain symmetry plane; and extracting the brain symmetry plane based on the target weight vector and the coordinates of the support vector points.

[0011] In another embodiment, the sampling point coordinates are input into a brain symmetry plane extraction model to perform a symmetry plane extraction operation, and obtaining the target weight vector, target bias and support vector point coordinates includes: in response to the sampling point coordinates, the initial weight vector and the initial bias satisfying constraints, the sampling point coordinates are input into a brain symmetry plane extraction model to perform a symmetry plane extraction operation so that the square of the norm of the initial weight vector is minimized to obtain the target weight vector, the target bias and the support vector point coordinates.

[0012] In another embodiment, extracting the brain symmetry plane based on the coordinates of the target weight vector and the support vector point includes: determining the direction and the center point of the brain symmetry plane based on the coordinates of the target weight vector and the support vector point, respectively; and extracting the brain symmetry plane according to the direction and the center point of the brain symmetry plane.

[0013] In yet another embodiment, the brain symmetry plane extraction model comprises a SVM model.

[0014] In a second aspect, the present application provides a device for extracting brain symmetry planes, comprising: a processor; and a memory, in which program instructions for extracting brain symmetry planes are stored, and when the program instructions are executed by the processor, the device implements multiple embodiments of the aforementioned first aspect.

[0015] In a third aspect, the present application provides a computer-readable storage medium having stored thereon computer-readable instructions for extracting brain symmetry planes, wherein when the computer-readable instructions are executed by one or more processors, multiple embodiments of the aforementioned first aspect are implemented.

[0016] Through the scheme for extracting brain symmetry planes as provided above, the embodiment of the present application first obtains brain partitioning results by performing brain partitioning based on cranial brain image data through a trained brain partitioning network model, and then performs symmetry plane extraction operations by sampling spatial coordinate points in the left brain area and the right brain area respectively, and inputting the sampling point coordinates into the brain symmetry plane extraction model to extract the brain symmetry plane. Based on this, the embodiment of the present application provides a more robust symmetry axis / symmetry plane extraction method through artificial intelligence, which can quickly and accurately segment the symmetry planes of the brain on normal and abnormal cranial brain images. Furthermore, the brain symmetry plane extraction model of the embodiment of the present application adopts the SVM model, which greatly improves the classification accuracy and makes the extracted brain symmetry plane more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0018] Figure 1 is an exemplary schematic diagram showing a method for extracting a brain symmetry plane according to an embodiment of the present application;

[0019] Figure 2 is an exemplary flowchart showing the operation of training a brain partition network model according to an embodiment of the present application;

[0020] Figure 3 is an exemplary schematic diagram showing a brain partition network model according to an embodiment of the present application;

[0021] Figure 4 is an exemplary schematic diagram showing extraction of brain symmetry planes according to an embodiment of the present application;

[0022] Figure 5 is an exemplary schematic diagram showing the extracted brain symmetry plane and brain partitioning results according to an embodiment of the present application;

[0023] Figure 6 It is an exemplary structural block diagram showing a device for extracting brain symmetry planes according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0025] It should be understood that the terms "include" and "comprising" used in the specification and claims of the present application indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0026] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this application specification and claims, unless the context clearly indicates otherwise, the singular forms of "a", "an" and "the" are intended to include plural forms. It should also be further understood that the term "and / or" used in this application specification and claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0027] As used in this specification and claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0028] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0029] Figure 1 FIG. 1 is an exemplary schematic diagram showing a method 100 for extracting a brain symmetry plane according to an embodiment of the present application. Figure 1As shown in, at step S101, the cranial brain image data of the brain symmetry plane to be extracted is input into the trained brain partition network model to perform brain partition operation to obtain the brain partition result. In an implementation scenario, the aforementioned cranial brain image data may include but is not limited to cranial brain CT image data and / or cranial brain MR image data, and all are three-dimensional tomographic data. In an implementation scenario, the brain partition network model can be trained by the following operations to obtain a trained brain partition network model. That is, first, the cranial brain image data of the marked left and right brain regions are divided into a training set, a validation set, and a test set, and the brain partition network model is trained using the training set to obtain the target partition result, and then the brain partition network model is verified and tested using the validation set and the test set to obtain a trained brain partition network model.

[0030] In some embodiments, the cranial brain imaging data can be labeled with left and right brain region labels according to the brain anatomical structure. As an example, 0-5 can be used to represent the left brain region, and 6-10 can be used to represent the right brain region. Based on the cranial brain imaging data labeled with the left and right brain regions, it is divided into a training set, a validation set, and a test set, and the training set is input into the brain partition network model for training to obtain the target partition result. Specifically, the brain partition network model is trained by optimizing the loss function between the target partition result and the true partition. The aforementioned loss function can be expressed as in Indicates the target partition result, S indicates the true partition, and Dice indicates the Dice loss. In addition, before the cranial brain imaging data is input into the brain partition network model, the brain partition network model can be processed in blocks. Then, the brain partition network model is verified and tested by the validation set and the test set to obtain a trained brain partition network model, and then the brain partition result is obtained by performing brain partition operations in the trained brain partition network model.

[0031] In one embodiment, the trained brain partition network model can be, for example, a U-net model, which can include an encoder and a decoder, and the encoder performs downsampling and outputs features related to brain partitioning, and the decoder uses the features output by the encoder as input, performs upsampling and outputs brain partitioning results. That is to say, by inputting cranial brain imaging data into the brain partition network model, the classification result is output after passing through the encoder, and the feature output of the encoder can be simultaneously input into the decoder, which finally outputs the segmentation result of the block area and maps it back to the corresponding position of the original image to obtain the brain partitioning result.

[0032] In some implementation scenarios, the encoder may include multiple layers of convolution blocks. For the convolution blocks of the encoder, each convolution block has a convolution operation and a pooling operation, and after the convolution operation, a dropout layer, a normalization layer, and an activation layer may be connected. Among them, the aforementioned pooling operation may be, for example, maximum pooling ("max pool") or average pooling ("average pool"), etc. In this scenario, each time the feature undergoes a pooling operation, the size of the feature map is reduced to half, so as to implement a downsampling operation and output the feature to expand it into a one-dimensional vector. In other implementation scenarios, the decoder may include multiple layers of deconvolution blocks. Among them, the input of the decoder is the feature of the output of the last layer of the encoder, and each time the feature undergoes a deconvolution operation, the size of its feature map is doubled, thereby implementing an upsampling operation. In addition, the embodiment of the present application also connects the output of the convolution layer of the same layer as the encoder with the deconvolution result of the previous layer in the decoder, and obtains multiple feature maps after multiple deconvolutions. Further, after convolution and softmax activation function, a brain partition result map with the same size as the input cranial brain imaging data is obtained. Will be combined later Figure 3 Describe the aforementioned brain partition network model in detail.

[0033] After obtaining the above-mentioned brain partitioning results, at step S102, spatial coordinate point sampling is performed in the left brain area and the right brain area respectively according to the brain partitioning results to obtain the corresponding sampling point coordinates. That is, sampling points are collected in the left and right brain areas output by the brain partitioning network model and the sampling point coordinates are obtained. Then, at step S103, the sampling point coordinates are input into the brain symmetry plane extraction model to perform a symmetry plane extraction operation to extract the brain symmetry plane. Preferably, the aforementioned brain symmetry plane extraction model can be, for example, a support vector machine (Support Vector Machines, "SVM") model. Specifically, in an implementation scenario, the sampling point coordinates are input into the brain symmetry plane extraction model to perform a symmetry plane extraction operation to obtain the target weight vector, target bias and support vector point coordinates. Among them, the support vector point is a pair of sampling points with the smallest midpoint distance on both sides of the left and right brain areas and the largest distance from the brain symmetry plane. Then, the brain symmetry plane is extracted based on the target weight vector and the coordinates of the support vector points.

[0034] More specifically, in one implementation scenario, in response to the sampling point coordinates, the initial weight vector, and the initial bias satisfying the constraint conditions, the sampling point coordinates are input into the brain symmetry plane extraction model to perform a symmetry plane extraction operation, so that the square of the norm of the initial weight vector is minimized to obtain the target weight vector, the target bias, and the coordinates of the support vector points. Further, the direction and center point of the brain symmetry plane are determined based on the coordinates of the target weight vector and the support vector points, and then the brain symmetry plane is extracted according to the direction and center point of the brain symmetry plane.

[0035] It can be understood that the SVM model is a supervised learning model for classification and regression analysis, which can find a hyperplane that can divide different categories in the feature space while maximizing the interval from the data point to the hyperplane so that the data points of two different categories are on both sides of the hyperplane. In an embodiment of the present application, the aforementioned different categories are the left brain area and the right brain area, and the maximum interval hyperplane is the brain symmetry plane. The maximum interval hyperplane refers to the plane with the maximum distance from the nearest sampling points of the two categories, and the two nearest sampling points are called support vector points. In an exemplary scenario, the aforementioned hyperplane can be expressed as:

[0036] w·x+b=0 (1)

[0037] Where w represents the weight vector, i.e., the normal vector of the hyperplane, x represents the input feature vector, i.e., the sampling points on both sides of the left and right brain regions, and b represents the bias. For a two-dimensional feature space, the aforementioned hyperplane is a straight line; in a three-dimensional space, the aforementioned hyperplane is a plane; in a higher-dimensional space, it is a hyperplane.

[0038] In an embodiment of the present application, the problem of finding the maximum margin hyperplane is transformed into a convex quadratic programming problem, that is, minimizing the square of the norm of the initial weight vector. Among them, the normal vector w of the maximum margin hyperplane is the weight in the direction of the hyperplane, which determines the importance of the feature. In an exemplary scenario, the aforementioned minimization of the square of the norm of the initial weight vector can be expressed by the following formula:

[0039]

[0040] Specifically, in the process of performing the symmetry plane extraction operation through the brain symmetry plane extraction model, when the input sampling point coordinates on both sides of the left and right brain regions, the initial weight vector and the initial bias meet the constraints, the square of the norm of the initial weight vector is minimized to obtain the coordinates of the target weight vector, the target bias and the support vector point. Among them, the aforementioned constraints are expressed as y i (w·x i +b)≥1,i=1,2,...,N,y i Represents the category label of the sampling point, such as 0 (representing the left brain area) and 1 (representing the right brain area). Further, the direction and center point of the brain symmetry plane are determined according to the coordinates of the target weight vector and the support vector points, thereby obtaining the brain symmetry plane. Preferably, the direction of the brain symmetry plane is determined according to the target weight vector (i.e., the normal vector -w[0] / w[2], -w[1] / w[2], -1) by taking the mean of the coordinate points of the support vector as the center point of the brain symmetry plane.

[0041] Combined with the above description, it can be seen that the embodiment of the present application obtains brain partitioning results by performing brain partitioning based on cranial brain image data through a trained brain partitioning network model, and then performs a symmetry plane extraction operation by inputting the sampling point coordinates of the left brain area and the right brain area into the brain symmetry plane extraction model to extract the brain symmetry plane. Preferably, the embodiment of the present application greatly improves the classification accuracy by adopting the SVM model, making the symmetry plane extraction method of the present application more robust, so that the symmetry plane of the brain can be segmented quickly and accurately on normal and abnormal cranial brain images.

[0042] Figure 2 FIG. 2 is an exemplary flowchart showing operation 200 of training a brain partition network model according to an embodiment of the present application. Figure 2 As shown in, at step S201, cranial brain image data are collected. In an implementation scenario, the aforementioned cranial brain image data may include but are not limited to cranial brain CT image data and / or cranial brain MR image data, and all are three-dimensional tomographic data. Then, at step S202, the cranial brain image data is labeled with left and right brain partitions. Based on the cranial brain image data of the labeled left and right brain regions, it is divided into a training set, a validation set, and a test set, and at step S203, the training set is used to train the brain partition network model to obtain a trained brain partition network model. Furthermore, the brain partition result is obtained by performing a brain partition operation in the trained brain partition network model.

[0043] As mentioned above, the trained brain partition network model can be, for example, a U-net model, and the brain partition network model can include an encoder and a decoder. The encoder can include multiple layers of convolution blocks, each of which has a convolution operation and a pooling operation. After the convolution operation, a dropout layer, a normalization layer, and an activation layer can be connected; the decoder can include multiple layers of deconvolution blocks. In the implementation scenario, by inputting cranial imaging data into the brain partition network model, the classification result is output after passing through the encoder, and the feature output of the encoder can be simultaneously input into the decoder, and the decoder finally outputs the brain partition result, for example Figure 3 shown.

[0044] Figure 3 FIG. 1 is an exemplary schematic diagram showing a brain partition network model according to an embodiment of the present application. Figure 3As shown in , the brain partition network model may include an encoder 301 on the left side and a decoder 302 on the right side. The encoder 301 may include multiple layers (e.g., 4 layers) of convolution blocks 303, and the decoder 302 may include multiple layers (e.g., 4 layers) of deconvolution blocks 304. For the encoder 301, each convolution block 303 may include a convolution operation (e.g., convolution + relu) and a pooling operation (e.g., max pool or average pool, etc.), and a 3*3 convolution kernel, for example, may be used in the convolution operation, and each layer of the convolution block may use, for example, 64, 128, 256, 612 convolution kernels, respectively. Then, after the convolution operation, a dropout operation, a normalization operation, an activation operation, etc. may be performed. As an example, assuming that the step size of the pooling operation is 2, after each pooling operation, the size of the feature map is reduced to half to achieve a downsampling operation and output the feature to expand it into a one-dimensional vector.

[0045] For the decoder 302, the input of the decoder is the feature of the output of the last layer of the encoder, such as the feature 305 shown in the figure. As an example, assuming that the step size of the deconvolution operation is 2, the size of the feature map doubles after each deconvolution operation, thereby realizing the upsampling operation. In addition, the output of the same convolution layer as the encoder is connected with the deconvolution result of the previous layer in the decoder, and multiple feature maps are obtained after multiple deconvolutions. For example, the output 306 of the same convolution layer as the encoder is connected with the deconvolution result 307 of the previous layer in the decoder. Similarly, multiple feature maps are obtained after multiple deconvolutions. In an exemplary scenario, assuming that the deconvolution block uses 256, 128, and 64 convolution kernels respectively, 64 feature maps can be obtained after three deconvolutions. Further, after another convolution of, for example, 32 convolution kernels and a softmax activation function, a brain partition result map 309 of the same size as the input cranial brain image data 308 is obtained.

[0046] After obtaining the above brain partition results, spatial coordinate point sampling is performed in the left brain area and the right brain area respectively to obtain the corresponding sampling point coordinates, and then the sampling point coordinates are input into the brain symmetry plane extraction model to perform symmetry plane extraction operations to extract the brain symmetry plane. In an implementation scenario, the aforementioned brain symmetry plane extraction model can be, for example, an SVM model. Specifically, the sampling point coordinates, the initial weight vector, and the initial bias satisfy the constraint conditions (i.e., y i (w·x i +b)≥1, i=1, 2, ..., N), the sampling point coordinates are input into the brain symmetry plane extraction model to perform symmetry plane extraction operations so that the square of the norm of the initial weight vector is minimized (i.e. ) to obtain the coordinates of the target weight vector, target bias and support vector points to obtain the maximum margin hyperplane, which is the brain symmetry plane of the embodiment of the present application. That is, the distance from the two nearest sampling points of the left brain area and the right brain area is the largest, for example Figure 4 shown.

[0047] Figure 4 FIG. 2 is an exemplary schematic diagram showing the extraction of brain symmetry planes according to an embodiment of the present application. Figure 4 The solid line in represents the location of the hyperplane L, the circle shown on the left side of the hyperplane L represents the sampling points of the left brain region, and the small triangle shown on the right side of the hyperplane L represents the sampling points of the right brain region. In an embodiment of the present application, by inputting the sampling points of the left and right brain regions into, for example, an SVM model, under constraints, the square of the norm of the initial weight vector is minimized to obtain the target weight vector, the target bias, and the coordinates of the support vector points. Among them, the support vector points are the pairs of sampling points with the smallest distance between the midpoints on both sides of the left and right brain regions and the largest distance from the brain symmetry plane, such as a1 and a2 shown in the figure. Furthermore, by taking the mean of the coordinate points of the support vector as the center point of the brain symmetry plane, the direction of the brain symmetry plane is determined according to the target weight vector to obtain the brain symmetry plane.

[0048] Figure 5 FIG. 1 is an exemplary schematic diagram showing the extracted brain symmetry plane and brain partitioning results according to an embodiment of the present application. Figure 5 The first column shows the extraction results of different cranial brain image data using the traditional method in the horizontal direction. The positions indicated by the straight lines in the first column are the extracted symmetry surfaces. Figure 5 The second column shows the extraction results of the embodiment of the present application in the horizontal direction for different cranial brain image data, and the positions indicated by the straight lines in each figure in the second column are the extracted symmetry planes. Figure 5 The third column shows the brain partitioning results extracted by the brain partitioning network model for different cranial brain image data. Different grayscales in the figure represent different functional areas of the brain, and the left and right brain areas are labeled after brain partitioning. Figure 5 The first column uses the traditional method to extract the symmetry plane, and the symmetry plane obtained by the present application solution is more accurate.

[0049] Figure 6 is an exemplary structural block diagram showing an apparatus 600 for extracting brain symmetry planes according to an embodiment of the present application.

[0050] like Figure 6As shown in , the device of the present application may include a central processing unit or central processing unit ("CPU") 611, which may be a general-purpose CPU, a dedicated CPU, or other information processing and program execution unit. Further, the device 600 may also include a large-capacity memory 612 and a read-only memory ("ROM") 613, wherein the large-capacity memory 612 may be configured to store various types of data, including various cranial and brain imaging data, brain partition results, sampling point coordinates, brain symmetry planes, algorithm data, intermediate results, and various programs required to run the device 600. The ROM 613 may be configured to store the power-on self-test of the device 600, the initialization of each functional module in the system, the basic input / output driver of the system, and the data and instructions required to boot the operating system.

[0051] Optionally, the device 600 may also include other hardware platforms or components, such as the tensor processing unit ("TPU") 614, graphics processing unit ("GPU") 615, field programmable gate array ("FPGA") 616, and machine learning unit ("MLU") 617 shown. It is understood that although a variety of hardware platforms or components are shown in the device 600, they are merely exemplary and not restrictive, and those skilled in the art may add or remove corresponding hardware according to actual needs. For example, the device 600 may include only a CPU, related storage devices, and interface devices to implement the method for extracting brain symmetry planes of the present application.

[0052] In some embodiments, in order to facilitate the transmission and interaction of data with an external network, the device 600 of the present application also includes a communication interface 618, so that it can be connected to a local area network / wireless local area network ("LAN / WLAN") 605 through the communication interface 618, and then connected to a local server 606 or to the Internet ("Internet") 607 through the LAN / WLAN. Alternatively or additionally, the device 600 of the present application can also be directly connected to the Internet or a cellular network through the communication interface 618 based on wireless communication technology, such as wireless communication technology based on the third generation ("3G"), the fourth generation ("4G") or the fifth generation ("5G") of the generation. In some application scenarios, the device 600 of the present application can also access the server 608 and the database 609 of the external network as needed to obtain various known algorithms, data and modules, and can remotely store various data, such as various types of data or instructions for presenting cranial brain imaging data, brain partition results, sampling point coordinates, brain symmetry planes, etc.

[0053] The peripheral devices of the device 600 may include a display device 602, an input device 603 and a data transmission interface 604. In one embodiment, the display device 602 may include, for example, one or more speakers and / or one or more visual displays, which are configured to perform voice prompts and / or image video display for extracting brain symmetry planes of the present application. The input device 603 may include, for example, other input buttons or controls such as a keyboard, a mouse, a microphone, a gesture capture camera, etc., which are configured to receive input and / or user instructions of audio data. The data transmission interface 604 may include, for example, a serial interface, a parallel interface or a universal serial bus interface ("USB"), a small computer system interface ("SCSI"), a serial ATA, a firewire ("FireWire"), a PCI Express and a high-definition multimedia interface ("HDMI"), etc., which are configured to transmit and interact with data of other devices or systems. According to the scheme of the present application, the data transmission interface 604 can receive cranial CT image data and cranial MR image data collected from a CT image acquisition device or an MR image acquisition device, and transmit to the device 600 data or results including cranial CT image data, cranial MR image data or various other types of cranial CT image data, cranial MR image data or various other types of cranial MR image data.

[0054] The CPU 611, mass storage 612, ROM 613, TPU 614, GPU 615, FPGA 616, MLU 617 and communication interface 618 of the device 600 of the present application can be interconnected through a bus 619, and data interaction with peripheral devices can be achieved through the bus. In one embodiment, through the bus 619, the CPU 611 can control other hardware components in the device 600 and its peripheral devices.

[0055] Combination of the above Figure 6 The device for extracting brain symmetry planes that can be used to implement the present application is described. It should be understood that the device structure or architecture here is only exemplary, and the implementation method and implementation entity of the present application are not limited thereto, but can be changed without departing from the spirit of the present application.

[0056] According to the above description in combination with the accompanying drawings, those skilled in the art can also understand that the embodiments of the present application can also be implemented by software programs. Therefore, the present application also provides a computer-readable storage medium on which computer-readable instructions for extracting brain symmetry planes are stored. When the computer-readable instructions are executed by one or more processors, the present application in combination with the accompanying drawings can be implemented. Figure 1 The described method for extracting brain symmetry planes.

[0057] It should be noted that although the operations of the method of the present application are described in a specific order in the accompanying drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. On the contrary, the steps depicted in the flow chart can be performed in a different order. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step, and / or one step can be decomposed into multiple steps.

[0058] It should be understood that when the terms "first", "second", "third" and "fourth" are used in the claims, the specification and the drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0059] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this application specification and claims, unless the context clearly indicates otherwise, the singular forms of "a", "an" and "the" are intended to include plural forms. It should also be further understood that the term "and / or" used in this application specification and claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0060] Although the implementation methods of the present application are as above, the contents described are only examples adopted to facilitate the understanding of the present application, and are not intended to limit the scope and application scenarios of the present application. Any technician in the technical field described in the present application can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present application, but the scope of patent protection of the present application shall still be subject to the scope defined in the attached claims.

Claims

1. A method for extracting brain symmetry planes, comprising: Inputting the cranial brain image data of the brain symmetry plane to be extracted into the trained brain partitioning network model to perform brain partitioning operation to obtain the brain partitioning result; According to the brain partitioning result, spatial coordinate point sampling is performed in the left brain area and the right brain area respectively to obtain corresponding sampling point coordinates; as well as The sampling point coordinates are input into the brain symmetry plane extraction model to perform a symmetry plane extraction operation to extract the brain symmetry plane. The sampling point coordinates are input into the brain symmetry plane extraction model to perform a symmetry plane extraction operation to extract the brain symmetry plane, including: Inputting the sampling point coordinates into a brain symmetry plane extraction model to perform a symmetry plane extraction operation, and obtaining the coordinates of a target weight vector, a target bias, and support vector points, wherein the support vector points are pairs of sampling points with the smallest midpoint distance between the left and right brain regions and the largest distance from the brain symmetry plane; and The brain symmetry plane is extracted based on the target weight vector and the coordinates of the support vector points, wherein the target weight vector is the normal vector of the brain symmetry plane. The sampling point coordinates are input into the brain symmetry plane extraction model to perform symmetry plane extraction operations, and the coordinates of the target weight vector, target bias and support vector points are obtained, including: In response to the sampling point coordinates, the initial weight vector and the initial bias satisfying the constraint conditions, the sampling point coordinates are input into the brain symmetry plane extraction model to perform a symmetry plane extraction operation so that the square of the norm of the initial weight vector is minimized to obtain the target weight vector, the target bias and the coordinates of the support vector point.

2. The method according to claim 1, wherein the brain partition network model is trained by the following operations to obtain the trained brain partition network model: Divide the cranial imaging data with labeled left and right brain regions into training set, validation set and test set; Using the training set to train the brain partition network model to obtain a target partition result; as well as The brain partition network model is verified and tested using the verification set and the test set to obtain the trained brain partition network model.

3. The method according to claim 1 or 2, wherein the trained brain partition network model includes an encoder and a decoder, and the encoder performs downsampling and outputs features related to brain partition, and the decoder takes the features output by the encoder as input, performs upsampling and outputs the brain partition result.

4. The method according to claim 1 or 2, wherein the brain image data at least includes brain CT image data and / or brain MR image data.

5. The method according to claim 1, wherein extracting the brain symmetry plane based on the coordinates of the target weight vector and the support vector points comprises: Determine the direction and center point of the brain symmetry plane based on the target weight vector and the coordinates of the support vector points respectively; as well as The brain symmetry plane is extracted according to the direction and the center point of the brain symmetry plane. The method according to claim 1 , wherein the brain symmetry plane extraction model comprises a SVM model.

7. A device for extracting brain symmetry planes, comprising: processor; as well as A memory having program instructions for extracting brain symmetry planes stored therein, wherein when the program instructions are executed by the processor, the device implements the method according to any one of claims 1-6.

8. A computer-readable storage medium storing computer-readable instructions for extracting brain symmetry planes, wherein the computer-readable instructions, when executed by one or more processors, implement the method according to any one of claims 1 to 6.

Images