Cerebral hemorrhage surgical robot and electronic equipment

Abstract

The embodiment of the invention provides a cerebral hemorrhage surgical robot and electronic equipment. The cerebral hemorrhage surgical robot comprises a control host and a mechanical arm, a preprocessing module, an intracranial hematoma prediction model and a post-processing module are deployed on the control host, and the preprocessing module is used for preprocessing a to-be-processed brain axial fault grayscale image so as to convert the image into an input image of an intracranial hematoma segmentation module; the intracranial hematoma prediction model is used for performing feature extraction and voxel classification on the input image to obtain a prediction result of intracranial hematoma; and the post-processing module is used for performing three-dimensional reconstruction on the intracranial hematoma according to the prediction result of the intracranial hematoma to obtain an intracranial hematoma model so as to control the action of the mechanical arm.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of robot technology, and in particular to a cerebral hemorrhage surgery robot and electronic equipment. Background Art

[0002] Currently, the reconstruction techniques of intracranial hematoma include threshold segmentation method, region growing-based method and level set-based method.

[0003] The threshold segmentation method aims to effectively separate the hematoma area from the image by accurately setting the grayscale threshold. However, the implementation of this method is highly dependent on manual operation, and the threshold must be adjusted frequently to ensure the best separation effect. In addition, since the grayscale value of the bone in the image is similar to that of the hematoma, this method may mistakenly include the bone in the image while separating the hematoma, and the bone needs to be manually removed.

[0004] The method based on region growing starts from a specific point in the image and gradually absorbs surrounding points with similar grayscale to form and continuously expand the hematoma area until no new points that meet the conditions are added. The use of this method requires the user to manually locate the hematoma first and specify the starting point of growth. For complex situations with multiple bleeding sites, the user needs to perform multiple operations to complete the reconstruction.

[0005] The level set-based method requires the user to set the initial contour, and with the help of the evolution and iteration process of the level set function, gradually adjust the contour shape and position, and finally achieve accurate segmentation of the target area. However, this method has high computational complexity and relatively slow iteration speed, which may affect the overall reconstruction efficiency.

[0006] In addition, the above three methods all focus on the overall reconstruction of intracranial hematoma and fail to achieve more refined granularity control, which to some extent limits their application effect in complex cases. Summary of the invention

[0007] In view of this, an embodiment of the present invention provides a cerebral hemorrhage surgery robot and an electronic device to at least partially solve the above problems.

[0008] According to a first aspect of an embodiment of the present invention, a brain hemorrhage surgical robot is provided, which includes: a control host and a robotic arm, wherein a preprocessing module, an intracranial hematoma prediction model, and a post-processing module are deployed on the control host, wherein the preprocessing module is used to preprocess the brain axial tomographic grayscale image to be processed to convert it into an input image of the intracranial hematoma segmentation module; the intracranial hematoma prediction model is used to perform feature extraction and voxel classification on the input image to obtain a prediction result of the intracranial hematoma; the post-processing module is used to perform three-dimensional reconstruction of the intracranial hematoma according to the prediction result of the intracranial hematoma to obtain an intracranial hematoma model to control the movement of the robotic arm.

[0009] Optionally, when the preprocessing module preprocesses the grayscale image of the brain axial section to be processed to convert it into an input image of the intracranial hematoma segmentation module, the following steps are performed:

[0010] Based on the total number of input channels in the intracranial hematoma segmentation module, the to-be-processed brain axial tomographic grayscale image is processed in blocks to obtain a plurality of brain axial tomographic grayscale image sub-blocks, so that the total number of voxels included in each brain axial tomographic grayscale image sub-block is equal to the total number of input channels;

[0011] Perform center enhancement processing on each brain axial grayscale image sub-block to obtain an influence matrix map;

[0012] Each brain axial tomographic grayscale image sub-block is enhanced according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block, and the fused brain axial tomographic grayscale image sub-block is used as the input image.

[0013] Optionally, the preprocessing module performs block processing on the brain axial tomographic grayscale image to be processed based on the total number of input channels in the intracranial hematoma segmentation module to obtain a plurality of brain axial tomographic grayscale image sub-blocks, and executes the following steps:

[0014] Calculating an X-direction block parameter for performing block processing on the brain axial tomography grayscale image to be processed along the X direction based on a difference between the size of the brain axial tomography grayscale image to be processed in the X direction and the planned size of a single brain axial tomography grayscale image sub-block in the X direction;

[0015] Based on the size of the brain axial tomographic grayscale image to be processed in the Y direction and the difference between the planned sizes of a single brain axial tomographic grayscale image sub-block in the Y direction, calculating the Y-direction block parameters when the brain axial tomographic grayscale image to be processed is block processed along the Y direction;

[0016] Based on the size of the to-be-processed brain axial tomographic grayscale image in the Z direction and the difference between the planned sizes of a single brain axial tomographic grayscale image sub-block in the Z direction, calculating the Z-direction block parameters when the to-be-processed brain axial tomographic grayscale image is block processed along the Z direction;

[0017] Based on the X-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the X-direction to obtain a plurality of X-direction tomographic grayscale image sub-blocks;

[0018] Based on the Y-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the Y-direction to obtain a plurality of Y-direction tomographic grayscale image sub-blocks;

[0019] Based on the Z-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the Z-direction to obtain a plurality of Z-direction tomographic grayscale image sub-blocks;

[0020] Based on the plurality of X-direction tomographic grayscale image sub-blocks, the plurality of Y-direction tomographic grayscale image sub-blocks, and the plurality of Z-direction tomographic grayscale image sub-blocks, the plurality of brain axial tomographic grayscale image sub-blocks are generated.

[0021] Optionally, the preprocessing module performs center enhancement processing on each brain axial tomographic grayscale image sub-block to obtain an influence matrix diagram, and executes the following steps:

[0022] determining the dimension of performing center enhancement processing on each brain axial tomographic grayscale image sub-block, so as to determine the direction of performing center enhancement processing based on the number of the dimensions;

[0023] Based on the direction of the center enhancement processing, influence enhancement processing is performed on a plurality of voxels in the center area of ​​each brain axial tomographic grayscale image sub-block along the dimension to obtain an influence matrix diagram.

[0024] Optionally, the preprocessing module performs the following steps when enhancing each brain axial tomographic grayscale image sub-block according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block:

[0025] The grayscale values ​​of a plurality of voxels in the central area of ​​each brain axial tomographic grayscale image sub-block are positively adjusted according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block.

[0026] Optionally, the intracranial hematoma prediction model is trained according to the following steps to perform feature extraction and voxel classification on the input image to obtain a prediction result of intracranial hematoma:

[0027] Acquire a brain axial tomographic grayscale image sample and a corresponding intracranial hematoma label, wherein the intracranial hematoma label indicates that the category of the brain axial tomographic grayscale image is background, brain parenchymal hemorrhage, or intraventricular hemorrhage;

[0028] Preprocessing the brain axial tomographic grayscale image samples to obtain preprocessed image samples;

[0029] Inputting the preprocessed image samples into the specified neural network model for feature extraction and voxel classification to calculate the predicted probability of each intracranial hematoma label;

[0030] Calculating the loss value of the neural network model according to the predicted probability of the intracranial hematoma label;

[0031] The network parameters of the neural network model are adjusted based on the loss value, so that the neural network model when the loss value reaches a set loss threshold is used as the intracranial hematoma prediction model.

[0032] Optionally, when the intracranial hematoma prediction model is trained, the brain axial tomographic grayscale image samples are preprocessed to obtain preprocessed image samples, including:

[0033] The voxels of the brain axial grayscale image samples are corrected for spatial consistency and grayscale value distribution is normalized to obtain preprocessed image samples.

[0034] Optionally, when the intracranial hematoma prediction model is trained, the preprocessed image samples are input into the specified neural network model for feature extraction and voxel classification to calculate the prediction probability of each intracranial hematoma label, including:

[0035] Inputting the preprocessed image sample into the feature extraction module in the specified neural network model to extract features to obtain feature vector samples;

[0036] The feature vector samples are input into the output module in the neural network model for three-category prediction to obtain three-channel four-dimensional output data, wherein the first dimension of the three-channel four-dimensional data includes the length information of the voxels in the preprocessed image samples, the second dimension includes the width information of the voxels in the preprocessed image samples, the third dimension includes the width information of the voxels in the preprocessed image samples, and the fourth dimension includes three channels, and one channel represents a predicted probability sample of an intracranial hematoma label.

[0037] Optionally, when the preprocessing module preprocesses the brain axial tomographic grayscale image to be processed to convert it into an input image of the intracranial hematoma segmentation module, each brain axial tomographic grayscale image sub-block is preprocessed to convert it into an input image of the intracranial hematoma segmentation module, and one brain axial tomographic grayscale image sub-block corresponds to one input image;

[0038] The intracranial hematoma prediction model performs feature extraction and voxel classification on the input image according to the following steps to obtain a prediction result of the intracranial hematoma. One input image corresponds to one prediction result of the intracranial hematoma:

[0039] The feature extraction module in the intracranial hematoma prediction model extracts features from the input image to obtain a feature vector;

[0040] The feature extraction module in the intracranial hematoma prediction model performs three-category prediction based on the feature vector to obtain three-channel four-dimensional output data, wherein the first dimension data in the three-channel four-dimensional data includes the length information of the voxels in the input image, the second dimension data includes the width information of the voxels in the input image, the third dimension data includes the height information of the voxels in the input image, and the fourth dimension data includes three channels, one channel represents the predicted probability of an intracranial hematoma type, and the intracranial hematoma types are background, cerebral parenchymal hemorrhage, and ventricular hemorrhage, wherein the intracranial hematoma type corresponding to the channel with the largest predicted probability is used as the prediction result of the intracranial hematoma corresponding to the input image.

[0041] Optionally, the preprocessing module is also used to flip the input image along at least one of the X direction, Y direction, and Z direction of the brain axial tomographic grayscale image to be processed to obtain a flipped input image, so that the intracranial hematoma prediction model performs feature extraction on the flipped input image to obtain a prediction sub-result of the intracranial hematoma, and the prediction sub-results of the intracranial hematoma corresponding to all the flipped input features are combined to obtain a prediction result of the intracranial hematoma corresponding to the input image.

[0042] Optionally, the post-processing module performs the following steps when performing three-dimensional reconstruction of the intracranial hematoma according to the prediction result of the intracranial hematoma to obtain the intracranial hematoma model:

[0043] splicing prediction results of intracranial hematoma corresponding to all input images to obtain a global prediction result of intracranial hematoma corresponding to the brain axial tomographic grayscale image to be processed;

[0044] splicing the influence matrix maps corresponding to all brain axial tomographic grayscale image sub-blocks to obtain a global influence matrix map corresponding to the brain axial tomographic grayscale image to be processed;

[0045] Based on the global influence matrix diagram of the to-be-processed brain axial tomographic grayscale image, a three-dimensional weighted adjustment is performed on the global prediction result of the corresponding intracranial hematoma to obtain an output prediction result of the intracranial hematoma;

[0046] Restoring the output prediction result of the intracranial hematoma into a three-dimensional grayscale image;

[0047] Isosurfaces are extracted from the three-dimensional grayscale image, and all extracted isosurfaces are spliced ​​to generate an intracranial hematoma model.

[0048] According to a second aspect of an embodiment of the present invention, there is provided an electronic device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; the memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform the following steps:

[0049] Preprocessing the grayscale image of the brain axial section to be processed to convert it into an input image of the intracranial hematoma segmentation module;

[0050] Performing feature extraction and voxel classification on the input image to obtain a prediction result of intracranial hematoma;

[0051] A three-dimensional reconstruction of the intracranial hematoma is performed according to the prediction result of the intracranial hematoma to obtain an intracranial hematoma model to control the movement of the mechanical arm.

[0052] In the scheme of the embodiment of the present invention, the brain hemorrhage surgical robot automatically processes the grayscale image of the brain axial section through the preprocessing module, and uses the intracranial hematoma prediction model to extract features, without the need for frequent manual adjustment of the threshold, greatly reducing manual intervention and improving the degree of automation. At the same time, the prediction model can more accurately identify the hematoma area through length learning and other technologies, effectively avoiding the mis-segmentation of similar grayscale value areas such as bones. In addition, the robot of the present invention can identify and segment all hematoma areas at one time through the global feature extraction ability of the intracranial hematoma prediction model, without the need for the user to manually locate multiple times, significantly improving the efficiency and accuracy of processing complex cases. Furthermore, the robot of the present invention adopts an efficient prediction model (such as a length learning network), which is usually optimized and can achieve a faster processing speed while ensuring high accuracy, thereby overcoming the computational efficiency problem of the level set method. Finally, the post-processing module of the present invention can perform high-precision three-dimensional reconstruction according to the prediction results, generate a detailed intracranial hematoma model, and realize the separate reconstruction of the intracranial hematoma model for intracranial brain parenchymal hemorrhage and ventricular hemorrhage. Especially in complex cases, this refined control is particularly important. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0054] Figure 1 This is a schematic diagram of a cerebral hemorrhage surgery robot according to an embodiment of the present application.

[0055] Figure 2 A schematic flow chart of a method for establishing an intracranial hematoma model provided in this embodiment.

[0056] Figure 3 It is a schematic diagram of the prediction result of intracranial hematoma according to one embodiment of the present invention.

[0057] Figure 4 Schematic diagram of an intracranial hematoma model according to an embodiment of the present invention. Figure 5 FIG. 1 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be described clearly and in detail below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in the field based on the embodiments in the embodiments of the present invention should fall within the scope of protection of the embodiments of the present invention.

[0059] It should be understood that the terms "first", "second", and "third" in the claims, specifications, and drawings of the present disclosure are used to distinguish different objects rather than to describe a specific order. The terms "include" and "comprise" used in the specifications and claims of the present disclosure 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.

[0060] It should also be understood that the terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. As used in this disclosure and claims, the singular forms of "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations.

[0061] Figure 1 Schematic diagram of a brain hemorrhage surgery robot according to an embodiment of the present application. Figure 1 As shown, it includes: a control host and a robotic arm, wherein a preprocessing module, an intracranial hematoma prediction model, and a post-processing module are deployed on the control host, wherein the preprocessing module is used to preprocess the grayscale image of the axial brain section to be processed so as to convert it into an input image of the intracranial hematoma segmentation module; the intracranial hematoma prediction model is used to perform feature extraction and voxel classification on the input image so as to obtain a prediction result of the intracranial hematoma; the post-processing module is used to perform three-dimensional reconstruction of the intracranial hematoma according to the prediction result of the intracranial hematoma, and obtain an intracranial hematoma model so as to control the movement of the robotic arm.

[0062] In the scheme of the embodiment of the present invention, the brain hemorrhage surgical robot automatically processes the grayscale image of the brain axial section through the preprocessing module, and uses the intracranial hematoma prediction model to extract features, without the need for frequent manual adjustment of the threshold, greatly reducing manual intervention and improving the degree of automation. At the same time, the prediction model can more accurately identify the hematoma area through length learning and other technologies, effectively avoiding the mis-segmentation of similar grayscale value areas such as bones. In addition, the robot of the present invention can identify and segment all hematoma areas at one time through the global feature extraction ability of the intracranial hematoma prediction model, without the need for the user to manually locate multiple times, significantly improving the efficiency and accuracy of processing complex cases. Furthermore, the robot of the present invention adopts an efficient prediction model (such as a length learning network), which is usually optimized and can achieve a faster processing speed while ensuring high accuracy, thereby overcoming the computational efficiency problem of the level set method. Finally, the post-processing module of the present invention can perform high-precision three-dimensional reconstruction according to the prediction results, generate a detailed intracranial hematoma model, and realize the separate reconstruction of the intracranial hematoma model for intracranial brain parenchymal hemorrhage and ventricular hemorrhage. Especially in complex cases, this refined control is particularly important.

[0063] Optionally, the brain axial tomographic grayscale image to be processed may be, but is not limited to, CT (Computed Tomography) or CTA (Computed Tomography Angiography).

[0064] Optionally, when the preprocessing module preprocesses the grayscale image of the brain axial section to be processed to convert it into an input image of the intracranial hematoma segmentation module, the following steps are performed:

[0065] Based on the total number of input channels in the intracranial hematoma segmentation module, the to-be-processed brain axial tomographic grayscale image is processed in blocks to obtain a plurality of brain axial tomographic grayscale image sub-blocks, so that the total number of voxels included in each brain axial tomographic grayscale image sub-block is equal to the total number of input channels;

[0066] Perform center enhancement processing on each brain axial grayscale image sub-block to obtain an influence matrix map;

[0067] Each brain axial tomographic grayscale image sub-block is enhanced according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block, and the fused brain axial tomographic grayscale image sub-block is used as the input image.

[0068] To this end, through the steps of block processing, center enhancement processing and enhancement based on the influence matrix, the following technical benefits are achieved:

[0069] 1. Improve data processing efficiency and adaptability

[0070] Block processing: This technical solution first blocks the grayscale image of the brain axial section to be processed according to the total number of input channels in the intracranial hematoma segmentation module. The advantage of this is that it can divide large-size images into multiple smaller blocks that are easier to process. This divide-and-conquer strategy can significantly improve data processing efficiency, especially when processing high-resolution images. In addition, block processing also enhances the adaptability of the system, enabling it to flexibly process images of different sizes.

[0071] 2. Enhance key area information

[0072] Center enhancement processing: Each brain axial tomography grayscale image sub-block is subjected to center enhancement processing to obtain an influence matrix. The purpose of this step is to highlight the key areas in the image, especially those that may be associated with intracranial hematoma. By enhancing the information of these areas, the accuracy of subsequent segmentation can be improved. As a weighted tool, the influence matrix can guide subsequent enhancement processing to ensure that important information receives sufficient attention.

[0073] 3. Improve segmentation accuracy and robustness

[0074] Enhancement based on influence matrix: Enhancement processing is performed according to the influence matrix corresponding to each brain axial tomography grayscale image sub-block to obtain enhanced brain axial tomography grayscale image sub-blocks. These enhanced sub-blocks are used as input images for the intracranial hematoma segmentation module, which can provide richer and more accurate information. Since the information in the key areas is enhanced, the segmentation module can be more accurate in identifying hematomas. At the same time, this enhancement processing also improves the robustness of the system, making it more stable in the face of interference factors such as noise and artifacts.

[0075] 4. Achieve refined segmentation

[0076] Fusion enhancement sub-blocks: Finally, all enhanced brain axial tomography grayscale image sub-blocks will be fused as the input image of the intracranial hematoma segmentation module. This process achieves a refined segmentation of the original image. Since each sub-block has been enhanced in a targeted manner, the segmentation accuracy of the entire image has been significantly improved. This refined segmentation is crucial for accurately identifying the morphology, location, and relationship between intracranial hematoma and other tissues.

[0077] Optionally, the preprocessing module performs block processing on the brain axial tomographic grayscale image to be processed based on the total number of input channels in the intracranial hematoma segmentation module to obtain a plurality of brain axial tomographic grayscale image sub-blocks, and executes the following steps:

[0078] Calculating an X-direction block parameter for performing block processing on the brain axial tomography grayscale image to be processed along the X direction based on a difference between the size of the brain axial tomography grayscale image to be processed in the X direction and the planned size of a single brain axial tomography grayscale image sub-block in the X direction;

[0079] Based on the size of the brain axial tomographic grayscale image to be processed in the Y direction and the difference between the planned sizes of a single brain axial tomographic grayscale image sub-block in the Y direction, calculating the Y-direction block parameters when the brain axial tomographic grayscale image to be processed is block processed along the Y direction;

[0080] Based on the size of the to-be-processed brain axial tomographic grayscale image in the Z direction and the difference between the planned sizes of a single brain axial tomographic grayscale image sub-block in the Z direction, calculating the Z-direction block parameters when the to-be-processed brain axial tomographic grayscale image is block processed along the Z direction;

[0081] Based on the X-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the X-direction to obtain a plurality of X-direction tomographic grayscale image sub-blocks;

[0082] Based on the Y-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the Y-direction to obtain a plurality of Y-direction tomographic grayscale image sub-blocks;

[0083] Based on the Z-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the Z-direction to obtain a plurality of Z-direction tomographic grayscale image sub-blocks;

[0084] Based on the plurality of X-direction tomographic grayscale image sub-blocks, the plurality of Y-direction tomographic grayscale image sub-blocks, and the plurality of Z-direction tomographic grayscale image sub-blocks, the plurality of brain axial tomographic grayscale image sub-blocks are generated.

[0085] To this end, by accurately calculating the block parameters and performing block processing along the X, Y, and Z directions, the following technical benefits are achieved:

[0086] 1. Accurate block division to improve processing efficiency

[0087] This technical solution divides the grayscale image of the brain axial section into blocks by accurately calculating the block parameters in the three directions of X, Y, and Z. This precise block division method can ensure that the size of each sub-block is moderate, neither too large to cause low processing efficiency nor too small to cause information loss or increase processing complexity. Therefore, this block processing can significantly improve the efficiency of image processing.

[0088] 2. Strong adaptability and flexible processing of images of different sizes

[0089] Since the technical solution calculates the block parameters based on the difference between the size of the image to be processed and the planned size of a single sub-block, it can flexibly adapt to images of different sizes. No matter how the image size changes, the appropriate sub-block size can be obtained by adjusting the block parameters, thereby ensuring the accuracy and stability of image processing.

[0090] 3. Maintain image integrity and avoid information loss

[0091] In traditional block processing, improper block division may lead to loss or duplication of image information. However, this technical solution can ensure that each sub-block contains complete image information by accurately calculating the block division parameters and performing block division along the three directions of X, Y, and Z, thus avoiding loss or duplication of information. This is crucial for subsequent image processing and segmentation.

[0092] 4. Provide convenience for subsequent processing

[0093] The multiple brain axial grayscale image sub-blocks obtained after block processing can be used as input data for subsequent processing. These sub-blocks not only contain complete image information, but also have small size and moderate complexity, which facilitates subsequent image processing, feature extraction and segmentation.

[0094] Optionally, the preprocessing module performs center enhancement processing on each brain axial tomographic grayscale image sub-block to obtain an influence matrix diagram, and executes the following steps:

[0095] determining the dimension of performing center enhancement processing on each brain axial tomographic grayscale image sub-block, so as to determine the direction of performing center enhancement processing based on the number of the dimensions;

[0096] Based on the direction of the center enhancement processing, influence enhancement processing is performed on a plurality of voxels in the center area of ​​each brain axial tomographic grayscale image sub-block along the dimension to obtain an influence matrix diagram.

[0097] To this end, the center enhancement processing technology has the following technical benefits:

[0098] 1. Improve the prominence of key area information

[0099] Center enhancement processing: This technology can significantly highlight the information in the central area of ​​each brain axial grayscale image sub-block by performing center enhancement processing on each sub-block. In medical image processing, the central area often contains the most critical information, such as lesions, hematomas, etc. Through center enhancement processing, this information can be more easily identified and extracted in subsequent processing.

[0100] 2. Improve segmentation accuracy

[0101] Influence matrix: The center enhancement process not only enhances the information of the center area, but also generates an influence matrix. This matrix reflects the influence of each voxel on the center area, providing an important reference for subsequent segmentation processing. By using the influence matrix, the lesion boundary can be more accurately located and the segmentation accuracy can be improved.

[0102] 3. Enhance the robustness of the algorithm

[0103] Multi-dimensional processing: The technical solution mentions that the dimension and direction of the center enhancement processing can be determined as needed. This multi-dimensional processing method can adapt to lesions of different shapes and directions and enhance the robustness of the algorithm. Even if the shape and position of the lesion change, the technology can ensure accurate extraction of information by adjusting the processing dimension and direction.

[0104] 4. Improve processing efficiency

[0105] Targeted enhancement: The center enhancement process only enhances the central area of ​​the sub-block, avoiding indiscriminate processing of the entire sub-block. This targeted enhancement method can reduce the amount of calculation and improve processing efficiency. At the same time, since only the information in the key area is enhanced, the subsequent processing is more focused and efficient.

[0106] 5. Provide convenience for subsequent processing

[0107] Preprocessing optimization: As one of the preprocessing steps, center enhancement processing provides optimized input data for subsequent feature extraction, segmentation, etc. These optimized data have higher signal-to-noise ratio and better separability, which helps the subsequent processing algorithms to perform better.

[0108] Optionally, the preprocessing module performs the following steps when enhancing each brain axial tomographic grayscale image sub-block according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block:

[0109] The grayscale values ​​of a plurality of voxels in the central area of ​​each brain axial tomographic grayscale image sub-block are positively adjusted according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block.

[0110] Therefore, the technical solution of enhancing the sub-blocks of the brain axial tomographic grayscale image according to the influence matrix diagram has the following technical advantages:

[0111] 1. Highlight key information

[0112] By positively adjusting the voxel grayscale values ​​in the central area of ​​the sub-block of the brain axial tomography grayscale image through the influence matrix, the information of these key areas can be significantly highlighted. In medical image processing, especially for the segmentation of intracranial hematoma, the central area often contains the most important lesion information. By enhancing these areas, doctors or subsequent processing algorithms can more easily identify and locate lesions.

[0113] 2. Improve segmentation accuracy

[0114] The enhanced brain axial tomographic grayscale image sub-blocks have higher signal-to-noise ratio and contrast, which makes the boundary between the lesion and the surrounding tissue clearer. For the subsequent segmentation algorithm, this means that the lesion boundary can be identified more accurately, thereby improving the segmentation accuracy. This is crucial for the precise operation of the cerebral hemorrhage surgical robot.

[0115] 3. Adaptive Enhancement

[0116] The influence matrix is ​​generated based on the specific situation of each sub-block, so the enhancement process is adaptive. This adaptive enhancement ensures that appropriate enhancement effects can be obtained between different sub-blocks and between different areas within sub-blocks. This avoids over-enhancement or under-enhancement, thereby improving the overall processing quality.

[0117] 4. Maintain image coherence

[0118] Although only the central area of ​​the sub-block is enhanced, the enhancement is smooth and coherent due to the guidance of the influence matrix map. This avoids abrupt boundaries or artifacts during the enhancement process, thereby maintaining the overall coherence and authenticity of the image.

[0119] 5. Provide optimized input for subsequent processing

[0120] The enhanced brain axial grayscale image sub-blocks, as input data for subsequent processing, have higher quality and richer information. This means that the required information can be extracted more accurately for subsequent processing steps such as feature extraction, classification, and recognition, thereby improving the performance and accuracy of the entire processing flow.

[0121] Optionally, the intracranial hematoma prediction model is trained according to the following steps to perform feature extraction and voxel classification on the input image to obtain a prediction result of intracranial hematoma:

[0122] Acquire a brain axial tomographic grayscale image sample and a corresponding intracranial hematoma label, wherein the intracranial hematoma label indicates that the category of the brain axial tomographic grayscale image is background, brain parenchymal hemorrhage, or intraventricular hemorrhage;

[0123] Preprocessing the brain axial tomographic grayscale image samples to obtain preprocessed image samples;

[0124] Inputting the preprocessed image samples into the specified neural network model for feature extraction and voxel classification to calculate the predicted probability of each intracranial hematoma label;

[0125] Calculating the loss value of the neural network model according to the predicted probability of the intracranial hematoma label;

[0126] The network parameters of the neural network model are adjusted based on the loss value, so that the neural network model when the loss value reaches a set loss threshold is used as the intracranial hematoma prediction model.

[0127] In summary, the above training process of the intracranial hematoma prediction model provides strong support for the accurate prediction and auxiliary diagnosis of cerebral hemorrhage surgical robots, and its technical benefits are as follows:

[0128] 1. Improve forecast accuracy

[0129] Samples and labels: By obtaining brain axial grayscale tomographic image samples and corresponding intracranial hematoma labels, the training process can clearly distinguish between different categories such as background, brain parenchymal hemorrhage, and intraventricular hemorrhage. This clear classification helps improve the accuracy of the model in prediction and ensures the correct identification of intracranial hematoma.

[0130] Feature extraction: Using the specified neural network model to extract features from the preprocessed image samples can deeply mine the key information in the image, such as the shape, size, and location of the hematoma, thereby further improving the accuracy of the prediction.

[0131] 2. Optimize model performance

[0132] Loss value calculation: The loss value of the neural network model is calculated based on the predicted probability of the intracranial hematoma label, which can quantify the error of the model in the prediction process. This quantification helps to evaluate the performance of the model and provide a basis for subsequent network parameter adjustments.

[0133] Parameter adjustment: Adjusting the network parameters of the neural network model based on the loss value can gradually optimize the performance of the model. Through continuous iterative training and parameter adjustment, the loss value of the model can be gradually reduced to reach the set loss threshold, thereby obtaining better prediction results.

[0134] 3. Improve training efficiency

[0135] Preprocessing step: Preprocessing the brain axial grayscale image samples, such as block segmentation and center enhancement, can reduce the noise and redundant information in the image and improve the image quality. This helps to speed up the training of the neural network and improve the training efficiency.

[0136] Iterative training: Through continuous iterative training and parameter adjustment, the model can gradually converge to the optimal solution. This iterative training method can make full use of the information in the data set and improve the generalization ability and prediction performance of the model.

[0137] 4. Enhance the robustness of the model

[0138] Multi-sample training: Using a large number of brain axial grayscale image samples for training enables the model to learn more hematoma features in different situations. This helps to enhance the robustness of the model, enabling it to give accurate prediction results when faced with intracranial hematomas of different shapes, sizes and locations.

[0139] Loss threshold setting: By setting the loss threshold to control the model training process, we can ensure that the model stops training after reaching a certain performance level. This helps to avoid overfitting or underfitting of the model, thereby maintaining the stability and robustness of the model.

[0140] When the model is trained as above, the ReLu originally used in the prior art as the activation function is replaced by Leaky Relu as the activation function:

[0141] The expression of the leaky ReLU function is

[0142]

[0143] Among them, α is selected as 0.01. Compared with the ReLU function, the leaky ReLU function no longer outputs 0 for the negative part, but outputs a very small negative number. This can solve the problem that some neurons in the model will not be activated during training, resulting in the inability to update the weights.

[0144] In addition, for the loss function, the sum of the Dice loss function and the cross entropy loss function is selected as the loss function:

[0145] The Dice loss function is:

[0146] Where X i represents the true label of the i-th category (also called ground truth), Y i Represents the result of predicting the i-th category.

[0147] The cross entropy loss function is: L CE =-∑y(i)logP(i)

[0148] Where y(i) represents the value of the i-th category in the true label, which is 0 or 1, and P(i) represents the probability of predicting the i-th category.

[0149] To this end, the combination of Dice loss function and cross entropy loss function can fully utilize the advantages of both and further improve the performance of the model. The specific technical benefits are as follows:

[0150] (1) By optimizing the Dice loss and the cross entropy loss simultaneously, the model can focus on the overlap between the prediction results and the true labels while also considering the accuracy of the overall prediction distribution, thereby achieving more comprehensive optimization.

[0151] (2) Improving model stability: The combination of the robustness of the Dice loss function to imbalanced data and the fast convergence of the cross entropy loss function helps to improve the stability of the model during training and reduce the risk of overfitting or underfitting.

[0152] (3) Adapting to different task requirements: Different tasks may have different requirements for model performance. Combining Dice loss and cross entropy loss can make the model more flexible and able to adapt to the needs of different tasks. For example, in image segmentation tasks, the segmentation accuracy and classification effect can be balanced by adjusting the weights of the two.

[0153] Optionally, when the intracranial hematoma prediction model is trained, the brain axial tomographic grayscale image samples are preprocessed to obtain preprocessed image samples, including:

[0154] The voxels of the brain axial grayscale image samples are corrected for spatial consistency and grayscale value distribution is normalized to obtain preprocessed image samples.

[0155] Therefore, preprocessing the brain axial grayscale image samples (including spatial consistency correction and grayscale value distribution normalization) has the following technical benefits in the training of intracranial hematoma prediction model:

[0156] 1. Improve image quality

[0157] Spatial consistency correction: Due to factors such as imaging equipment, scanning parameters or patient position, there may be spatial inconsistencies in the voxels in the brain axial tomographic grayscale image samples. Through spatial consistency correction, these inconsistencies can be eliminated, making the voxels in the image more accurate and consistent in space. This helps to improve the quality of the image and provide a more reliable basis for subsequent feature extraction and prediction.

[0158] Grayscale value distribution standardization: The grayscale value distribution of different image samples may be different, which will affect the model's recognition and extraction of features. Through grayscale value distribution standardization, the grayscale values ​​of different image samples can be unified into the same range, eliminating the differences in grayscale value distribution. This helps the model better learn the features in the image and improve the accuracy of prediction.

[0159] 2. Enhance model generalization ability

[0160] Reducing data bias: The preprocessing step can reduce the data bias in the image samples, so that the model can focus more on the essential features of the image during training, rather than being distracted by irrelevant factors. This helps to enhance the generalization ability of the model, so that it can give accurate prediction results when faced with image samples from different sources and of different quality.

[0161] 3. Improve training efficiency

[0162] Accelerate model convergence: Through the preprocessing step, noise and redundant information in the image samples can be eliminated, so that the model can focus more on learning useful features during training. This helps to accelerate the convergence of the model, reduce training time, and improve training efficiency.

[0163] 4. Facilitate subsequent processing

[0164] Unify image format: The preprocessing step can unify image samples of different formats and resolutions into the same format and resolution, which is convenient for subsequent processing steps such as feature extraction, classification and recognition. This helps to simplify the entire processing flow and improve processing efficiency and accuracy.

[0165] When performing grayscale value distribution standardization, the tissue density value (such as HU value, Hounsfield Unit) can also be standardized. For example, since there are often abnormally large and small isolated values ​​in CT images, narrowing the range of HU values ​​to within the [0.5, 99.5] percentile interval can effectively remove interference. For this reason, the corresponding relationship between [0.5, 99.5] and grayscale value is: 0.5 percentile grayscale value T 0.5 , 99.5 percentile gray value T 99.5 .

[0166] To this end, after the gray value distribution is standardized, the gray value of each voxel satisfies the following relationship:

[0167]

[0168] Where M is the grayscale image sample of the axial tomography of the brain, which has a total of p voxels. The grayscale image sample of the axial tomography of the brain after the grayscale value distribution is standardized is recorded as N, and the voxels are recorded as N p, σ is the standard deviation of the voxel grayscale values ​​in the brain axial tomography grayscale image samples, and μ is the mean of the voxel grayscale values ​​in the brain axial tomography grayscale image samples.

[0169] Optionally, when the intracranial hematoma prediction model is trained, the preprocessed image samples are input into the specified neural network model for feature extraction and voxel classification to calculate the prediction probability of each intracranial hematoma label, including:

[0170] Inputting the preprocessed image sample into the feature extraction module in the specified neural network model to extract features to obtain feature vector samples;

[0171] The feature vector samples are input into the output module of the neural network model for three-category prediction to obtain three-channel four-dimensional output data, wherein the first dimension of the three-channel four-dimensional data includes the length information of the voxels in the preprocessed image samples, the second dimension includes the width information of the voxels in the preprocessed image samples, the third dimension includes the height information of the voxels in the preprocessed image samples, and the fourth dimension includes three channels, and one channel represents a predicted probability sample of an intracranial hematoma label.

[0172] Therefore, when training the intracranial hematoma prediction model, inputting the preprocessed image samples into the specified neural network model for feature extraction and calculation of the prediction probability has the following technical benefits:

[0173] 1. Efficient feature extraction

[0174] Feature extraction module: Through the feature extraction module in the neural network model, the preprocessed image samples can be efficiently extracted to obtain feature vector samples. This step can deeply mine the key information in the image, such as the shape, size, location of the hematoma, and the contrast with the surrounding tissues, etc., providing a strong basis for subsequent predictions.

[0175] 2. Accurate prediction

[0176] Three-classification prediction: The feature vector samples are input into the output module of the neural network model for three-classification prediction, and three-channel four-dimensional output data can be obtained. This design ensures that the model makes independent predictions for each voxel and gives the prediction probabilities of three intracranial hematoma labels (background, brain parenchymal hemorrhage, and intraventricular hemorrhage). This helps to improve the accuracy of the prediction and ensure the accurate identification of intracranial hematoma.

[0177] 3. Rich output information

[0178] Three-dimensional data structure: The output three-channel four-dimensional data has rich information, where the first dimension data contains the length information of the voxel, the second dimension data includes the width information of the voxel in the preprocessed image sample, the third dimension data includes the height information of the voxel in the preprocessed image sample, and the fourth dimension data includes three channels, each channel represents a predicted probability sample of an intracranial hematoma label. This data structure can fully reflect the situation of each voxel in the image and provide a detailed reference for doctors' diagnosis and treatment.

[0179] 4. Improve model performance

[0180] Independent prediction and comprehensive judgment: The model makes independent predictions for each voxel and combines the prediction results of all voxels to make the final judgment. This design can make full use of the information in the image and improve the performance and accuracy of the model. At the same time, by comparing the prediction probabilities of different voxels, the model can also make a more accurate assessment of the distribution and range of intracranial hematomas.

[0181] 5. Facilitate subsequent processing

[0182] Unified data format: The output three-channel four-dimensional data has a unified data format, which is convenient for subsequent processing and analysis. This helps to simplify the entire processing flow and improve processing efficiency and accuracy. At the same time, this data format is also convenient for integration and docking with other medical image processing systems to achieve data sharing and interoperability.

[0183] Optionally, when the preprocessing module preprocesses the brain axial tomographic grayscale image to be processed to convert it into an input image of the intracranial hematoma segmentation module, each brain axial tomographic grayscale image sub-block is preprocessed to convert it into an input image of the intracranial hematoma segmentation module, and one brain axial tomographic grayscale image sub-block corresponds to one input image;

[0184] The intracranial hematoma prediction model performs feature extraction and voxel classification on the input image according to the following steps to obtain a prediction result of the intracranial hematoma. One input image corresponds to one prediction result of the intracranial hematoma:

[0185] The feature extraction module in the intracranial hematoma prediction model extracts features from the input image to obtain a feature vector;

[0186] The feature extraction module in the intracranial hematoma prediction model performs three-category prediction based on the feature vector to obtain three-channel four-dimensional output data, wherein the first dimension data in the three-channel four-dimensional data includes the length information of the voxels in the input image, the second dimension data includes the width information of the voxels in the input image, the third dimension data includes the height information of the voxels in the input image, and the fourth dimension data includes three channels, one channel represents the predicted probability of an intracranial hematoma type, and the intracranial hematoma types are background, cerebral parenchymal hemorrhage, and ventricular hemorrhage, wherein the intracranial hematoma type corresponding to the channel with the largest predicted probability is used as the prediction result of the intracranial hematoma corresponding to the input image.

[0187] To this end, the above scheme for the preprocessing module and the intracranial hematoma prediction model in processing the brain axial tomographic grayscale image and obtaining the prediction result has the following technical advantages:

[0188] (1) Block processing improves efficiency:

[0189] The preprocessing module divides the grayscale image of the brain axial section to be processed into multiple sub-blocks and performs independent preprocessing on each sub-block. This block processing method can reduce the amount of calculation and improve the processing efficiency, especially when processing large-scale image data.

[0190] (2) Refined preprocessing:

[0191] By preprocessing each sub-block independently, more refined processing can be performed according to the characteristics of the sub-block, such as local enhancement, noise suppression, etc., thereby improving the image quality and the accuracy of subsequent processing.

[0192] (3) Efficiency of feature extraction module:

[0193] The feature extraction module in the intracranial hematoma prediction model can efficiently extract features from the input image and obtain a feature vector containing key information, which helps to reduce redundant information and improve the accuracy of subsequent predictions.

[0194] (4) Comprehensiveness of the three-category forecast:

[0195] The model performs three-category prediction and outputs three-channel four-dimensional data, taking into account the background, intracerebral hemorrhage and intraventricular hemorrhage. This comprehensive prediction method can ensure the accurate identification of intracranial hematoma and avoid missed diagnosis or misdiagnosis.

[0196] (5) Clarity of prediction results:

[0197] The model compares the prediction probabilities of the three channels and selects the intracranial hematoma type corresponding to the channel with the highest prediction probability as the final prediction result. This clear prediction method facilitates decision-making and judgment by doctors or subsequent processing algorithms.

[0198] (6) Structuring of output data:

[0199] The output three-channel four-dimensional data has a clear structure, including the length information, width information, height information of the voxel and the predicted probability of the three channels. This structured data is convenient for subsequent analysis and processing, such as visualization, statistical analysis, etc.

[0200] Optionally, the preprocessing module is also used to flip the input image along at least one of the X direction, Y direction, and Z direction of the brain axial tomographic grayscale image to be processed to obtain a flipped input image, so that the intracranial hematoma prediction model performs feature extraction on the flipped input image to obtain a prediction sub-result of the intracranial hematoma, and the prediction sub-results of the intracranial hematoma corresponding to all the flipped input features are combined to obtain a prediction result of the intracranial hematoma corresponding to the input image.

[0201] To this end, the preprocessing module is also used to perform a flip operation on the input image, which has the following technical benefits:

[0202] (1) Improve the robustness of the prediction model:

[0203] By flipping the input image in the X direction, Y direction, Z direction (or at least one of them), images from different perspectives can be obtained. This allows the intracranial hematoma prediction model to be exposed to more diverse data during the training process, thereby improving its ability to recognize hematomas of different perspectives and shapes and enhancing the robustness of the model.

[0204] (2) Reduce forecast bias:

[0205] Since image flipping can produce a mirror effect, for some symmetrical or nearly symmetrical hematomas, the flipped image may help the model capture its features more accurately, thereby reducing the prediction bias caused by perspective or morphological differences.

[0206] (3) Increase data diversity:

[0207] The flipping operation actually provides more training samples for the model without changing the essential characteristics of the hematoma. This data augmentation technique helps increase the diversity of training data, allowing the model to better learn the general characteristics of hematomas rather than overfitting to specific training samples.

[0208] (4) Improve forecast accuracy:

[0209] By combining the prediction sub-results corresponding to multiple flipped input images, more comprehensive prediction information can be obtained. This helps reduce the prediction error under a single perspective and improves the accuracy of the overall prediction. Especially when dealing with complex or difficult-to-judge hematomas, the prediction results from multiple perspectives can provide doctors with a more reliable reference.

[0210] (5) Facilitate model optimization and debugging:

[0211] As a simple data transformation method, flipping operation is convenient for debugging and verification during model training and optimization. By observing the prediction results under different flipping directions, it is easier to find problems and shortcomings of the model, so as to make targeted improvements and optimizations.

[0212] Optionally, the post-processing module performs the following steps when performing three-dimensional reconstruction of the intracranial hematoma according to the prediction result of the intracranial hematoma to obtain the intracranial hematoma model:

[0213] splicing prediction results of intracranial hematoma corresponding to all input images to obtain a global prediction result of intracranial hematoma corresponding to the brain axial tomographic grayscale image to be processed;

[0214] splicing the influence matrix maps corresponding to all brain axial tomographic grayscale image sub-blocks to obtain a global influence matrix map corresponding to the brain axial tomographic grayscale image to be processed;

[0215] Based on the global influence matrix diagram of the to-be-processed brain axial tomographic grayscale image, a three-dimensional weighted adjustment is performed on the global prediction result of the corresponding intracranial hematoma to obtain an output prediction result of the intracranial hematoma;

[0216] Restoring the output prediction result of the intracranial hematoma into a three-dimensional grayscale image;

[0217] Isosurfaces are extracted from the three-dimensional grayscale image, and all extracted isosurfaces are spliced ​​to generate an intracranial hematoma model.

[0218] To this end, the technical solution in which the post-processing module performs three-dimensional reconstruction of the intracranial hematoma according to the prediction result of the intracranial hematoma to obtain the intracranial hematoma model has the following technical advantages:

[0219] (1) Accuracy of global prediction results:

[0220] By splicing the intracranial hematoma prediction results corresponding to all input images, the post-processing module can generate a global intracranial hematoma prediction result corresponding to the grayscale image of the axial tomography of the brain to be processed. This global consideration helps to reduce local prediction errors and improve the accuracy of the overall prediction.

[0221] (2) Effective use of the influence matrix:

[0222] The influence matrix reflects the influence of each voxel on the prediction result. By splicing the influence matrix corresponding to all brain axial tomography grayscale image sub-blocks, the post-processing module can obtain the global influence matrix, so as to more accurately evaluate the contribution of different regions to intracranial hematoma prediction.

[0223] (3) Accuracy of three-dimensional weighted adjustment:

[0224] Based on the global influence matrix, the post-processing module performs three-dimensional weighted adjustment on the global prediction results of intracranial hematoma. This adjustment method can fully consider the influence of different regions, making the final output prediction results more accurate and reliable.

[0225] (4) Restoration of three-dimensional grayscale images:

[0226] Restoring the output prediction results of intracranial hematoma into a three-dimensional grayscale image helps doctors understand and analyze the prediction results more intuitively. At the same time, this restoration method also facilitates subsequent isosurface extraction and model generation.

[0227] (5) Generation of intracranial hematoma model:

[0228] By extracting isosurfaces from 3D grayscale images and stitching all extracted isosurfaces, the post-processing module can generate an intracranial hematoma model, which can accurately reflect the shape and location of intracranial hematoma and provide strong support for doctors' diagnosis and treatment.

[0229] (6) Assisted surgical planning:

[0230] The intracranial hematoma model can also provide important reference information for surgical planning. Doctors can formulate a more reasonable surgical plan based on the morphology and location of the model to improve the success rate and safety of the operation.

[0231] Figure 2 A schematic flow chart of a method for establishing an intracranial hematoma model provided in this embodiment. Figure 3 It is a schematic diagram of the prediction result of intracranial hematoma according to one embodiment of the present invention. Figure 4 Schematic diagram of an intracranial hematoma model according to an embodiment of the present invention.

[0232] like Figure 2-4 As shown, the method includes:

[0233] Preprocessing the grayscale image of the brain axial section to be processed to convert it into an input image of the intracranial hematoma segmentation module;

[0234] Performing feature extraction and voxel classification on the input image to obtain a prediction result of intracranial hematoma;

[0235] The intracranial hematoma is reconstructed three-dimensionally according to the prediction result of the intracranial hematoma to obtain an intracranial hematoma model.

[0236] In the method for establishing the intracranial hematoma model, the exemplary explanation of the above steps can be found in the above Figure 1 Description of Examples.

[0237] The embodiment of the present application also provides an electronic device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus; the memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform the following steps:

[0238] Preprocessing the grayscale image of the brain axial section to be processed to convert it into an input image of the intracranial hematoma segmentation module;

[0239] Performing feature extraction and voxel classification on the input image to obtain a prediction result of intracranial hematoma;

[0240] A three-dimensional reconstruction of the intracranial hematoma is performed according to the prediction result of the intracranial hematoma to obtain an intracranial hematoma model to control the movement of the mechanical arm.

[0241] In the electronic device, an exemplary explanation of the above steps can be found in the above Figure 1 Description of Examples.

[0242] The above intracranial hematoma prediction model is, for example, but not limited to, the nnUNet model. In order to increase the portability of the model, the intracranial hematoma prediction model can be trained in the pytorch environment. By converting the intracranial hematoma prediction model into the ONNX format and porting it to the C++ platform for use, it can be applied to a variety of processing platforms, improving the portability of the preset model. Among them, ONNX (Open Neural Network Exchange) is a multi-framework shared, open protocol neural network exchange format that defines a set of standard formats that are independent of the environment and platform, and is used to enhance the interactivity of various AI models.

[0243] Reference Figure 4 , shows a schematic structural diagram of an electronic device according to another embodiment of the present invention. The specific embodiment of the present invention does not limit the specific implementation of the electronic device.

[0244] like Figure 4 As shown, the electronic device may include: a processor (processor) 402 , a communication interface (Communications Interface) 404 , a memory (memory) 406 storing a program 410 , and a communication bus 408 .

[0245] The processor, the communication interface, and the memory communicate with each other via a communication bus. The communication interface is used to communicate with other electronic devices or servers. The processor is used to execute a program, and specifically can execute the relevant steps in the above method embodiment. Specifically, the program may include a program code, and the program code includes computer operation instructions.

[0246] The processor may be a CPU, or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the smart device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0247] The memory is used to store programs. The memory may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0248] The program can be specifically used to enable the processor to perform the following operations: segment the hematoma image by means of a sliding window to obtain a segmented image; calculate and obtain a Gaussian weight map corresponding to the segmented image; voxel-wise multiply the segmented image with the Gaussian weight map corresponding to the segmented image to obtain a weighted map of the segmented image; input the weighted map of the segmented image into a preset model for processing to obtain a probability map of the segmented image; fuse the probability map of the segmented image and the Gaussian weight map corresponding to the segmented image to obtain a reconstructed hematoma image.

[0249] The above implementation methods are only used to illustrate the embodiments of the present invention, and are not intended to limit the embodiments of the present invention. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the embodiments of the present invention. Therefore, all equivalent technical solutions also belong to the scope of the embodiments of the present invention, and the patent protection scope of the embodiments of the present invention should be defined by the claims. The systems, devices, modules or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions.

[0250] For the convenience of description, the above device is described as being divided into various units according to their functions. Of course, when implementing the present invention, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0251] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0252] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0253] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0254] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0255] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory. The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0256] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0257] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0258] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0259] The present invention may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific transactions or implement specific abstract data types. The present invention may also be practiced in distributed computing environments where transactions are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0260] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

Claims

1. A robot for cerebral hemorrhage surgery, characterized in that: include: A control host and a robotic arm, wherein a preprocessing module, an intracranial hematoma prediction model, and a post-processing module are deployed on the control host, wherein the preprocessing module is used to preprocess the grayscale image of the axial brain section to be processed so as to convert it into an input image of the intracranial hematoma segmentation module; the intracranial hematoma prediction model is used to perform feature extraction and voxel classification on the input image so as to obtain a prediction result of the intracranial hematoma; the post-processing module is used to perform three-dimensional reconstruction of the intracranial hematoma according to the prediction result of the intracranial hematoma so as to obtain an intracranial hematoma model so as to control the movement of the robotic arm.

2. The cerebral hemorrhage surgery robot according to claim 1, characterized in that: When the preprocessing module preprocesses the grayscale image of the brain axial section to be processed to convert it into an input image of the intracranial hematoma segmentation module, the following steps are performed: Based on the total number of input channels in the intracranial hematoma segmentation module, the to-be-processed brain axial tomographic grayscale image is processed in blocks to obtain a plurality of brain axial tomographic grayscale image sub-blocks, so that the total number of voxels included in each brain axial tomographic grayscale image sub-block is equal to the total number of input channels; Perform center enhancement processing on each brain axial grayscale image sub-block to obtain an influence matrix map; Each brain axial tomographic grayscale image sub-block is enhanced according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block, and the fused brain axial tomographic grayscale image sub-block is used as the input image.

3. The cerebral hemorrhage surgery robot according to claim 2, characterized in that: The preprocessing module performs block processing on the to-be-processed brain axial tomographic grayscale image based on the total number of input channels in the intracranial hematoma segmentation module to obtain a plurality of brain axial tomographic grayscale image sub-blocks, and executes the following steps: Calculating an X-direction block parameter for performing block processing on the brain axial tomography grayscale image to be processed along the X direction based on a difference between the size of the brain axial tomography grayscale image to be processed in the X direction and the planned size of a single brain axial tomography grayscale image sub-block in the X direction; Based on the size of the brain axial tomographic grayscale image to be processed in the Y direction and the difference between the planned sizes of a single brain axial tomographic grayscale image sub-block in the Y direction, calculating the Y-direction block parameters when the brain axial tomographic grayscale image to be processed is block processed along the Y direction; Based on the size of the to-be-processed brain axial tomographic grayscale image in the Z direction and the difference between the planned sizes of a single brain axial tomographic grayscale image sub-block in the Z direction, calculating the Z-direction block parameters when the to-be-processed brain axial tomographic grayscale image is block processed along the Z direction; Based on the X-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the X-direction to obtain a plurality of X-direction tomographic grayscale image sub-blocks; Based on the Y-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the Y-direction to obtain a plurality of Y-direction tomographic grayscale image sub-blocks; Based on the Z-direction block parameter, the brain axial tomographic grayscale image to be processed is subjected to block processing along the Z-direction to obtain a plurality of Z-direction tomographic grayscale image sub-blocks; Based on the plurality of X-direction tomographic grayscale image sub-blocks, the plurality of Y-direction tomographic grayscale image sub-blocks, and the plurality of Z-direction tomographic grayscale image sub-blocks, the plurality of brain axial tomographic grayscale image sub-blocks are generated.

4. The cerebral hemorrhage surgery robot according to claim 2, characterized in that: The preprocessing module performs center enhancement processing on each brain axial tomographic grayscale image sub-block to obtain an influence matrix diagram, and executes the following steps: determining the dimension of performing center enhancement processing on each brain axial tomographic grayscale image sub-block, so as to determine the direction of performing center enhancement processing based on the number of the dimensions; Based on the direction of the center enhancement processing, influence enhancement processing is performed on a number of voxels in the central area of ​​each brain axial tomographic grayscale image sub-block along the dimension to obtain an influence matrix diagram.

5. The cerebral hemorrhage surgery robot according to claim 2, characterized in that: The preprocessing module performs the following steps when enhancing each brain axial tomographic grayscale image sub-block according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block: The grayscale values ​​of a plurality of voxels in the central area of ​​each brain axial tomographic grayscale image sub-block are positively adjusted according to the corresponding influence matrix diagram to obtain an enhanced brain axial tomographic grayscale image sub-block.

6. The cerebral hemorrhage surgery robot according to claim 1, characterized in that: The intracranial hematoma prediction model is trained according to the following steps to perform feature extraction and voxel classification on the input image to obtain a prediction result of intracranial hematoma: Acquire a brain axial tomographic grayscale image sample and a corresponding intracranial hematoma label, wherein the intracranial hematoma label indicates that the category of the brain axial tomographic grayscale image is background, brain parenchymal hemorrhage, or intraventricular hemorrhage; Preprocessing the brain axial tomographic grayscale image samples to obtain preprocessed image samples; Inputting the preprocessed image samples into the specified neural network model for feature extraction and voxel classification to calculate the predicted probability of each intracranial hematoma label; Calculating the loss value of the neural network model according to the predicted probability of the intracranial hematoma label; The network parameters of the neural network model are adjusted based on the loss value so that the neural network model when the loss value reaches a set loss threshold is used as the intracranial hematoma prediction model.

7. The cerebral hemorrhage surgery robot according to claim 6, characterized in that: When the intracranial hematoma prediction model is trained, the brain axial tomographic grayscale image samples are preprocessed to obtain preprocessed image samples, including: The voxels of the brain axial grayscale image samples are corrected for spatial consistency and grayscale value distribution is normalized to obtain preprocessed image samples.

8. The cerebral hemorrhage surgery robot according to claim 6, characterized in that: When the intracranial hematoma prediction model is trained, the preprocessed image samples are input into the specified neural network model for feature extraction and voxel classification to calculate the prediction probability of each intracranial hematoma label, including: Inputting the preprocessed image sample into the feature extraction module in the specified neural network model to extract features to obtain feature vector samples; The feature vector samples are input into the output module of the neural network model for three-category prediction to obtain three-channel four-dimensional output data, wherein the first dimension of the three-channel four-dimensional data includes the length information of the voxels in the preprocessed image samples, the second dimension includes the width information of the voxels in the preprocessed image samples, the third dimension includes the height information of the voxels in the preprocessed image samples, and the fourth dimension includes three channels, and one channel represents a predicted probability sample of an intracranial hematoma label.

9. The cerebral hemorrhage surgery robot according to claim 2, characterized in that: When the preprocessing module preprocesses the brain axial tomographic grayscale image to be processed to convert it into an input image of the intracranial hematoma segmentation module, each brain axial tomographic grayscale image sub-block is preprocessed to convert it into an input image of the intracranial hematoma segmentation module, and one brain axial tomographic grayscale image sub-block corresponds to one input image; The intracranial hematoma prediction model performs feature extraction and voxel classification on the input image according to the following steps to obtain a prediction result of the intracranial hematoma. One input image corresponds to one prediction result of the intracranial hematoma: The feature extraction module in the intracranial hematoma prediction model extracts features from the input image to obtain a feature vector; The feature extraction module in the intracranial hematoma prediction model performs three-category prediction based on the feature vector to obtain three-channel four-dimensional output data, wherein the first dimension data in the three-channel four-dimensional data includes the length information of the voxels in the input image, the second dimension data includes the width information of the voxels in the input image, the third dimension data includes the height information of the voxels in the input image, and the fourth dimension data includes three channels, one channel represents the predicted probability of an intracranial hematoma type, and the intracranial hematoma types are background, cerebral parenchymal hemorrhage, and ventricular hemorrhage, wherein the intracranial hematoma type corresponding to the channel with the largest predicted probability is used as the prediction result of the intracranial hematoma corresponding to the input image.

10. An electronic device comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform the following steps: Preprocessing the grayscale image of the brain axial section to be processed to convert it into an input image of the intracranial hematoma segmentation module; Performing feature extraction and voxel classification on the input image to obtain a prediction result of intracranial hematoma; A three-dimensional reconstruction of the intracranial hematoma is performed according to the prediction result of the intracranial hematoma to obtain an intracranial hematoma model to control the movement of the mechanical arm.

Images