Device for determining potential recurrence risk area of brain tumor and model training method

Abstract

The invention discloses a brain tumor potential recurrence risk area determination device and a model training method, and relates to the field of medical science and technology and the field of artificial intelligence. The device comprises a delineation determination unit used for determining a brain tumor delineation area based on a target medical image of a target object; the nerve fiber reconstruction unit is used for simulating and reconstructing image information of N target nerve fibers of the target object according to the target medical image; and the risk area prediction unit is used for inputting the target medical image and the image information of the N target nerve fibers into the target model, performing feature extraction on the received information through the target model, and determining a recurrence risk area of the brain tumor of the target object according to the extracted features by using priori knowledge in a model training stage. According to the method and the device, the technical problem of low accuracy of determining the potential recurrence risk area of the brain tumor caused by judging the risk diffusion condition of the brain tumor only through imaging observation or expert experience in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the fields of medical technology and artificial intelligence, and specifically, to a device for determining potential recurrence risk areas for brain tumors and a model training method. Background Art

[0002] High-grade gliomas are the most common and lethal human brain tumors. Even with standard treatment, these brain tumors are prone to recurrence due to their strong infiltrative ability, spreading through white matter fiber tracts to other brain regions.

[0003] White matter fiber tracts provide a pathway for tumor cells to migrate away from the primary tumor site, which is one of the reasons why complete tumor resection is difficult with surgery and localized treatments. White matter fiber tracts serve as pathways for the spread of malignant gliomas, promoting the widespread spread of tumor cells within the brain. Diffusion tensor imaging (DTI) is a highly sensitive imaging technique that can detect subtle interruptions in white matter fiber tracts.

[0004] However, in the existing technology, the risk spread of brain tumors is judged only based on diffusion tensor imaging through imaging observation or expert experience, and it is usually difficult to accurately determine the potential recurrence risk area of ​​brain tumors.

[0005] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0006] The present application provides a device for determining potential recurrence risk areas of brain tumors and a model training method to at least solve the technical problem in the prior art of low accuracy in determining potential recurrence risk areas of brain tumors, which results from judging the risk spread of brain tumors only through imaging observation or expert experience.

[0007] According to one aspect of the present application, a device for determining a potential recurrence risk area of ​​a brain tumor is provided, comprising: a delineation determination unit for determining a brain tumor delineation area of ​​a target object based on a target medical image of the target object; a nerve fiber reconstruction unit for simulating and reconstructing image information of N target nerve fibers of the target object based on the target medical image, wherein N is an integer greater than 1; and a risk area prediction unit for inputting the image information of the target medical image and the N target nerve fibers into a target model, performing feature extraction on the received information through the target model, and determining the recurrence risk area of ​​the brain tumor of the target object based on the extracted features using prior knowledge from the model training stage.

[0008] Optionally, the training data of the target model includes at least: the original onset area of ​​the brain tumor of the reference subject, the actual recurrence area of ​​the brain tumor, and the image information of the nerve fibers passing through the original onset area and the actual recurrence area.

[0009] Optionally, the device for determining the potential recurrence risk area of ​​brain tumors also includes: an image processing unit for reconstructing the image information of N nerve fibers of the reference object based on the medical image simulation of the reference object; a detection unit for detecting, based on the image information of the N nerve fibers of the reference object, whether each nerve fiber of the reference object simultaneously passes through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor of the reference object; a high-risk label setting unit for marking the nerve fibers that simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor as high-risk labels; a low-risk label setting unit for marking the nerve fibers that do not simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor as low-risk labels; and a model training unit for iteratively training the neural network based on the image information of the N nerve fibers of the reference object, the risk label marked with each nerve fiber, and the medical image of the reference object to obtain a target model.

[0010] Optionally, the delineation determination unit includes: a first delineation subunit, used to perform image delineation of the brain tumor area based on the medical image of the reference object taken when the disease first occurs, to obtain first delineation information; a second delineation subunit, used to expand the delineated area in the first delineation information by a preset proportion, to obtain second delineation information; and an original onset area determination subunit, used to determine the original onset area of ​​the brain tumor of the reference object according to the second delineation information.

[0011] Optionally, the delineation determination unit includes: a third delineation subunit, used to perform image delineation of the brain tumor area based on medical images taken of the reference object when the disease relapses, and obtain third delineation information; a recurrence area determination subunit, used to determine the actual recurrence area of ​​the brain tumor of the reference object according to the third delineation information.

[0012] Optionally, the risk area prediction unit includes: a label determination subunit, which is used to use the prior knowledge in the model training phase through the target model to determine the risk label carried by each target nerve fiber of the target object based on the extracted features, and treat the target nerve fibers carrying high-risk labels as high-risk nerve fibers; an image reconstruction subunit, which is used to reconstruct the high-risk nerve fibers into 3D images based on the coordinate system corresponding to the target object; an image combination subunit, which is used to combine all 3D images obtained based on the reconstruction of high-risk nerve fibers into a target 3D image; and a risk area determination subunit, which is used to determine the recurrence risk area of ​​the brain tumor of the target object based on the target 3D image.

[0013] Optionally, the risk area determination subunit includes: a detection module for detecting the density of high-risk nerve fibers in each area in the target 3D image; a rendering module for rendering the area where the density of high-risk nerve fibers is greater than a preset threshold as a highlighted area; and a determination module for determining the highlighted area as a recurrence risk area for the brain tumor of the target object.

[0014] Optionally, the third delineation subunit includes: a first image processing module, used to use a medical image of the reference subject taken before the recurrence of the disease as the first image; a second image processing module, used to use a medical image of the reference subject taken when the disease recurred as the second image; a registration module, used to register the first image and the second image to obtain a registration result; and a delineation module, used to perform image delineation of the brain tumor area according to the registration result to obtain third delineation information.

[0015] Optionally, the device for determining the potential recurrence risk area of ​​brain tumors also includes: an image acquisition unit, used to acquire a multimodal medical image of the target object; a data enhancement unit, used to perform data enhancement on the multimodal medical image of the target object, and use the multimodal medical image after data enhancement as the target medical image of the target object.

[0016] According to another aspect of the present application, a model training method is also provided, wherein the method includes: reconstructing image information of N nerve fibers of a reference object based on medical image simulation of the reference object; determining, from the N nerve fibers, the nerve fibers that simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor of the reference object; iteratively training a neural network based on the image information of the original onset area of ​​the brain tumor, the actual recurrence area of ​​the brain tumor, and the nerve fibers that simultaneously pass through the original onset area and the actual recurrence area of ​​the brain tumor of the reference object to obtain a target model.

[0017] In the device for determining the potential recurrence risk area of ​​a brain tumor of the present application, a delineation determination unit determines the brain tumor delineation area of ​​the target object based on the target medical image of the target object; a nerve fiber reconstruction unit simulates and reconstructs the image information of N target nerve fibers of the target object based on the target medical image, where N is an integer greater than 1; a risk area prediction unit inputs the target medical image and the image information of the N target nerve fibers into a target model, the target model performs feature extraction on the received information, and uses prior knowledge in the model training stage to determine the recurrence risk area of ​​the brain tumor of the target object based on the extracted features.

[0018] As can be seen from the above, the nerve fiber reconstruction unit of the present application uses the target medical image of the target object to simulate and reconstruct the image information of the nerve fibers of the target object, which can provide imaging evidence of the complex interaction between brain tumors and nerve fiber tissues. The visualization of nerve fibers helps to reveal the possible spread paths of brain tumors, which are difficult to effectively determine with traditional imaging observations. Finally, the present application also realizes the transformation from data to knowledge by inputting the target medical image and the reconstructed N nerve fiber image information into a neural network model (target model). Using the target model trained with a large amount of historical data, the key imaging features that affect the risk of recurrence can be automatically extracted, thereby predicting which areas have a higher risk of recurrence, and thus helping doctors to more accurately identify the brain areas that need to be focused on, which is of great reference value for formulating treatment plans such as surgical resection range and radiotherapy target area.

[0019] It can be seen that the technical solution of the present application solves the technical problem of low accuracy in determining the potential recurrence risk area of ​​brain tumors in the existing technology, which is caused by judging the risk spread of brain tumors only through imaging observation or expert experience, by introducing nerve fiber reconstruction and neural network models. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0021] Figure 1 is a schematic diagram of an optional device for determining a potential recurrence risk area of ​​a brain tumor according to an embodiment of the present application;

[0022] Figure 2 This is a flowchart of a model training method according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention 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. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] It should also be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) collected by this application are information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation portals for users to choose to authorize or refuse. For example, an interface is set up between this system and the relevant users or institutions. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or institution through the interface, and obtain the relevant information after receiving the consent information fed back by the aforementioned user or institution.

[0026] Figure 1 is a schematic diagram of an optional device for determining a potential recurrence risk area of ​​a brain tumor according to an embodiment of the present application, such as Figure 1 As shown, the apparatus for determining a potential recurrence risk area of ​​a brain tumor includes: a delineation determination unit 101 , a nerve fiber reconstruction unit 102 , and a risk area prediction unit 103 .

[0027] Optionally, the delineation determining unit 101 is configured to determine a brain tumor delineation region of the target object based on a target medical image of the target object.

[0028] Optionally, the target object may be a patient suffering from a brain tumor, and the target medical image may be an MRI image or other types of images, wherein the target medical image may be an image sequence, including but not limited to T1-weighted imaging (T1WI), enhanced T1-weighted imaging (CE-T1WI), T2-weighted imaging (T2WI), fluid-attenuated inversion recovery (FLAIR) imaging, diffusion-weighted imaging (DWI), and diffusion tensor imaging (DTI).

[0029] Optionally, after obtaining the target medical image of the target subject, the delineation determination unit 101 can use a neural network model to automatically delineate the target subject's brain tumor, thereby obtaining a delineated region of the target subject's brain tumor. The delineation determination unit 101 can also obtain a delineated region of the target subject's brain tumor by manual delineation. This application does not specifically limit the selection of the delineation method.

[0030] Optionally, the nerve fiber reconstruction unit 102 is configured to reconstruct image information of N target nerve fibers of the target object based on the target medical image simulation, where N is an integer greater than 1.

[0031] Optionally, there are multiple ways to reconstruct the nerve fibers, for example, reconstructing the nerve fibers through a reconstruction algorithm or reconstructing the nerve fibers through a neural network model.

[0032] The following is an optional nerve fiber reconstruction algorithm, including:

[0033] Step 1: Set seed points. Seed points can be automatically generated from the medical image. For example, seed points can be generated within the region of the medical image where the pixel value is greater than 0.7. An upper limit of 0.98 can also be set to remove possible noise. This means that seed points are generated within the region of the medical image where the pixel value is greater than 0.7 and less than 0.98.

[0034] Step 2: Construct a fiber response function to describe the diffusion movement of water molecules in biological tissues, and fit the seed point data and fiber response function into the constant solid angle direction distribution function model.

[0035] Step 3: Based on the constant solid angle direction distribution function model, calculate the sphere direction with the highest probability mass function. For example, the maximum angle can be set to 30 degrees, and the tracking and reconstruction of the nerve fibers is started, and finally N nerve fibers are reconstructed.

[0036] It should be noted that the above-mentioned nerve fiber reconstruction algorithm is only an optional embodiment. In actual operation, other methods can also be used to complete the reconstruction of nerve fibers. For example, a nerve fiber reconstruction model can be trained using medical images. The nerve fiber reconstruction model can be based on the image information of the medical images and use the prior knowledge related to nerve fibers learned by the model during the training phase to reconstruct the image information of the nerve fibers.

[0037] Optionally, the risk area prediction unit 103 is used to input the image information of the target medical image and N target nerve fibers into the target model, perform feature extraction on the received information through the target model, and use the prior knowledge in the model training stage to determine the recurrence risk area of ​​the target object's brain tumor based on the extracted features.

[0038] Optionally, the target model can be a pre-trained neural network model, and the training data of the target model includes at least: the original onset area of ​​the brain tumor of the reference subject, the actual recurrence area of ​​the brain tumor, and the image information of the nerve fibers passing through the original onset area and the actual recurrence area at the same time.

[0039] Among them, the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor of the reference subject can be represented in the form of outlined images. The neural network uses the outlined images of these two areas as training data. If a nerve fiber of the reference subject passes through the original onset area and the actual recurrence area at the same time, the nerve fiber is determined to be a high-risk nerve fiber; if a nerve fiber of the reference subject does not pass through the original onset area and the actual recurrence area at the same time, the nerve fiber is determined to be a low-risk nerve fiber. The training goal of the neural network is to accurately judge whether the nerve fiber of the reference subject is a high-risk nerve fiber based on the original onset area of ​​the brain tumor of the reference subject, the actual recurrence area of ​​the brain tumor, and the nerve fiber image information of the reference subject. After the training goal of the neural network is achieved, it can be determined that the target model has been obtained through training.

[0040] As can be seen from the above, the nerve fiber reconstruction unit of the present application uses the target medical image of the target object to simulate and reconstruct the image information of the nerve fibers of the target object, which can provide imaging evidence of the complex interaction between brain tumors and nerve fiber tissues. The visualization of nerve fibers helps to reveal the possible spread paths of brain tumors, which are difficult to effectively determine with traditional imaging observations. Finally, the present application also realizes the transformation from data to knowledge by inputting the target medical image and the reconstructed N nerve fiber image information into a neural network model (target model). Using the target model trained with a large amount of historical data, the key imaging features that affect the risk of recurrence can be automatically extracted, thereby predicting which areas have a higher risk of recurrence, and thus helping doctors to more accurately identify the brain areas that need to be focused on, which is of great reference value for formulating treatment plans such as surgical resection range and radiotherapy target area.

[0041] It can be seen that the technical solution of the present application solves the technical problem of low accuracy in determining the potential recurrence risk area of ​​brain tumors in the existing technology, which is caused by judging the risk spread of brain tumors only through imaging observation or expert experience, by introducing nerve fiber reconstruction and neural network models.

[0042] In an optional embodiment, the device for determining the potential recurrence risk area of ​​the brain tumor also includes: an image processing unit for reconstructing the image information of N nerve fibers of the reference object based on the medical image simulation of the reference object; a detection unit for detecting, based on the image information of the N nerve fibers of the reference object, whether each nerve fiber of the reference object simultaneously passes through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor of the reference object; a high-risk label setting unit for marking the nerve fibers that simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor as high-risk labels; a low-risk label setting unit for marking the nerve fibers that do not simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor as low-risk labels; and a model training unit for iteratively training the neural network based on the image information of the N nerve fibers of the reference object, the risk label marked with each nerve fiber, and the medical image of the reference object to obtain a target model.

[0043] Optionally, the image processing unit simulates and reconstructs image information of N nerve fibers in the brain of the reference object based on various medical image data (such as DTI data) of the reference object.

[0044] An optional implementation method is: first, extract the DTI sequence from the medical imaging data of the reference object, and then use the parameters such as anisotropic diffusion (AD), fractional anisotropy (FA), and geometric anisotropy (GA) in the DTI sequence, combined with the fiber response function and the constant solid angle direction distribution function (ODF) model to reconstruct the 3D image information of N nerve fibers.

[0045] Next, the detection unit checks whether each reconstructed nerve fiber passes through both the original onset area and the actual recurrence area of ​​the brain tumor. Based on the detection results of the detection unit, the high-risk label setting unit labels nerve fibers that pass through both the original onset area and the actual recurrence area as "high-risk." The low-risk label setting unit labels nerve fibers that do not pass through both the original onset area and the actual recurrence area of ​​the brain tumor as "low-risk."

[0046] Finally, the model training unit uses the image information of the labeled nerve fibers and the medical images of the reference object to train a neural network model, which can accurately determine whether each of the patient's nerve fibers is a high-risk nerve fiber based on the patient's reconstructed nerve fiber image information, the patient's original brain tumor onset area, and the actual brain tumor recurrence area. For example, the model training unit receives data from the image processing unit and the label setting unit and inputs this data into the neural network as a training set. Through iterative training, the neural network gradually adjusts the weights and parameters of each network layer of the neural network until the neural network can accurately identify which nerve fibers are high-risk nerve fibers and which nerve fibers are not high-risk nerve fibers. The neural network is then determined to have entered a convergence state and the neural network training is determined to be complete.

[0047] In an optional embodiment, the delineation determination unit includes: a first delineation subunit, used to perform image delineation of the brain tumor area based on the medical image of the reference object taken when the disease first occurs, to obtain first delineation information; a second delineation subunit, used to expand the delineated area in the first delineation information by a preset proportion, to obtain second delineation information; and an original disease area determination subunit, used to determine the original disease area of ​​the brain tumor of the reference object based on the second delineation information.

[0048] Optionally, assume that the medical image of the reference subject taken when the patient first became ill is image A. The first delineation subunit can support manual delineation or automatic delineation using a neural network model to delineate the brain tumor region in image A to obtain first delineation information (which can be recorded as delineation 1).

[0049] Then, the second delineation subunit can perform a certain degree of expansion on Delineation 1 to obtain second delineation information (which can be recorded as Delineation 2). The preset expansion ratio can be customized, for example, 20% expansion or 15% expansion. In addition, the direction and angle of the expansion can also be customized, which is not particularly limited in this embodiment of the application.

[0050] Finally, the original onset area of ​​the brain tumor of the reference subject was determined based on delineation 2.

[0051] It should be noted that the reason why the original onset area of ​​the brain tumor of the reference subject is determined based on Outline 2 is because there may be errors in Outline 1. By expanding outward by a certain proportion, the cancer cells on the boundary can be included in the original onset area of ​​the brain tumor as much as possible, thereby improving the accuracy of the subsequent judgment of the recurrence risk area.

[0052] Of course, the embodiment of the present application also supports directly using the outline 1 to determine the original onset area of ​​the brain tumor of the reference object.

[0053] In an optional embodiment, the delineation determination unit includes: a third delineation subunit, used to perform image delineation of the brain tumor area based on medical images taken of the reference subject when the disease relapses, to obtain third delineation information; and a recurrence area determination subunit, used to determine the actual recurrence area of ​​the brain tumor of the reference subject based on the third delineation information.

[0054] Optionally, assuming that the medical image of the reference subject taken at the time of recurrence is Image B, the third delineation subunit can support manual delineation or automatic delineation using a neural network model to delineate the brain tumor region in Image B, thereby obtaining third delineation information (referred to as Delineation 3). The recurrence region determination subunit can determine the actual recurrence region of the reference subject's brain tumor based on Delineation 3.

[0055] In an optional embodiment, the risk area prediction unit includes: a label determination subunit, which is used to use the prior knowledge in the model training stage through the target model to determine the risk label carried by each target nerve fiber of the target object based on the extracted features, and treat the target nerve fibers carrying high-risk labels as high-risk nerve fibers; an image reconstruction subunit, which is used to reconstruct the high-risk nerve fibers into 3D images based on the coordinate system corresponding to the target object; an image combination subunit, which is used to combine all 3D images obtained based on the reconstruction of the high-risk nerve fibers into a target 3D image; and a risk area determination subunit, which is used to determine the recurrence risk area of ​​the brain tumor of the target object based on the target 3D image.

[0056] Optionally, during the application stage of the target model, the image information of N target nerve fibers reconstructed from the target object and the target medical image data of the target object can be input into the target model. Then, through the prior knowledge learned by the target model during the model training stage, based on the relevant image features of the target nerve fibers and the relevant image features of the target medical images, the risk label carried by each target nerve fiber of the target object is determined, and the target nerve fibers carrying high-risk labels are treated as high-risk nerve fibers.

[0057] The image reconstruction subunit then reconstructs high-risk nerve fibers into 3D images based on the coordinate system corresponding to the target object. For example, for 100 target nerve fibers, 100 sets of 3D images are generated. These 100 sets of 3D images are then combined to create a target 3D image, which serves as an indicative image of high-risk recurrence areas specific to the target object.

[0058] In an optional embodiment, the risk area determination subunit includes: a detection module for detecting the density of high-risk nerve fibers in each area in the target 3D image; a rendering module for rendering the area where the density of high-risk nerve fibers is greater than a preset threshold as a highlighted area; and a determination module for determining the highlighted area as a recurrence risk area for the brain tumor of the target object.

[0059] Optionally, the detection module can detect the density of high-risk nerve fibers within each region of the target 3D image. A greater density of high-risk nerve fibers within a region indicates a higher likelihood of brain tumor recurrence. Based on this, the rendering module can highlight regions where the density of high-risk nerve fibers exceeds a preset threshold, alerting the physician that these regions are at risk of brain tumor recurrence. The preset threshold can be customized.

[0060] In an optional embodiment, the third delineation subunit includes: a first image processing module, configured to use a medical image of the reference subject taken before the recurrence of the disease as the first image; a second image processing module, configured to use a medical image of the reference subject taken when the disease recurred as the second image; a registration module, configured to register the first image and the second image to obtain a registration result; and a delineation module, configured to perform image delineation of the brain tumor area based on the registration result to obtain third delineation information.

[0061] Optionally, by registering the first image and the second image, the first image and the second image can be precisely aligned, thereby eliminating image discrepancies caused by factors such as head position and device calibration. The registration process includes, but is not limited to, various registration operations such as rigid body transformation, non-rigid body deformation, and feature point matching.

[0062] In an optional embodiment, the device for determining a potential recurrence risk area for a brain tumor further includes: an image acquisition unit for acquiring a multimodal medical image of a target object; and a data enhancement unit for performing data enhancement on the multimodal medical image of the target object and using the data-enhanced multimodal medical image as a target medical image of the target object.

[0063] Optionally, for various multimodal medical images such as DTI magnetic resonance images, data enhancement can be performed on these medical images through a data enhancement unit to obtain a multi-sequence image set as the target medical image, wherein the multi-sequence image set can include images of multiple modalities such as AD, FA, GA, linearity, MD, planarity, RD, sphericity and trace.

[0064] Among them, AD (Axial Diffusivity) is the axial diffusivity, which is used to reflect the diffusion rate of water molecules along the axis of nerve fibers and is particularly sensitive to detecting nerve fiber damage.

[0065] FA (Fractional Anisotropy) is fractional anisotropy, which is used to quantify the difference in the diffusion of water molecules in different directions. The value ranges from 0 to 1. A higher FA value indicates that the tissue has obvious diffusion directionality, which is usually associated with healthy white matter fibers.

[0066] GA (Geodesic Anisotropy) is geometric anisotropy, which is similar to FA but focuses on the curvature and bifurcation of fiber bundles. It is particularly useful for evaluating the geometric structural properties of fiber bundles.

[0067] MD (Mean Diffusivity) is the average diffusion rate, which provides information on the average diffusion speed of water molecules in all directions. MD is usually related to the overall health of the tissue.

[0068] RD (Radial Diffusivity) is the radial diffusivity, which is used to measure the diffusion rate of water molecules perpendicular to the axis of the nerve fiber. It is often used to study the integrity of the myelin sheath and the radial health of nerve fibers.

[0069] It should be noted that data augmentation can expand the number and diversity of images, thereby improving the robustness of the model and helping to improve the model's prediction accuracy for areas with potential recurrence risk of brain tumors.

[0070] According to another aspect of the embodiment of the present application, a model training method is also provided, wherein: Figure 2 is a flow chart of a model training method according to an embodiment of the present application, such as Figure 2 As shown, the method includes the following steps:

[0071] Step S201 : reconstructing image information of N nerve fibers of a reference object based on a medical image simulation of the reference object.

[0072] In step S202 , based on the image information of the N nerve fibers of the reference subject, the nerve fibers that simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor of the reference subject are determined from the N nerve fibers.

[0073] In step S203 , the neural network is iteratively trained based on the image information of the original onset area of ​​the brain tumor, the actual recurrence area of ​​the brain tumor, and the nerve fibers passing through the original onset area and the actual recurrence area of ​​the reference subject to obtain a target model.

[0074] Optionally, the target object may be a patient suffering from a brain tumor, and the target medical image may be an MRI image or other types of images, wherein the target medical image may be an image sequence, including but not limited to T1-weighted imaging (T1WI), enhanced T1-weighted imaging (CE-T1WI), T2-weighted imaging (T2WI), fluid-attenuated inversion recovery (FLAIR) imaging, diffusion-weighted imaging (DWI), and diffusion tensor imaging (DTI).

[0075] Optionally, after obtaining the target medical image of the target subject, a neural network model can be used to automatically delineate the target subject's brain tumor, thereby obtaining a delineated region of the target subject's brain tumor. Alternatively, manual delineation can be used to obtain the delineated region of the target subject's brain tumor. This application does not impose any particular limitation on the selection of the delineation method.

[0076] Optionally, there are multiple ways to reconstruct the nerve fibers, for example, reconstructing the nerve fibers through a reconstruction algorithm or reconstructing the nerve fibers through a neural network model.

[0077] The following is an optional nerve fiber reconstruction algorithm, including:

[0078] Step 1: Set seed points. Seed points can be automatically generated from the medical image. For example, seed points can be generated within the region of the medical image where the pixel value is greater than 0.7. An upper limit of 0.98 can also be set to remove possible noise. This means that seed points are generated within the region of the medical image where the pixel value is greater than 0.7 and less than 0.98.

[0079] Step 2: Construct a fiber response function to describe the diffusion movement of water molecules in biological tissues, and fit the seed point data and fiber response function into the constant solid angle direction distribution function model.

[0080] Step 3: Based on the constant solid angle direction distribution function model, calculate the direction of the sphere with the highest probability mass function, set the maximum angle to 30 degrees, start tracking and reconstructing the nerve fibers, and finally reconstruct N nerve fibers.

[0081] It should be noted that the above-mentioned nerve fiber reconstruction algorithm is only an optional embodiment. In actual operation, other methods can also be used to complete the reconstruction of nerve fibers. For example, a nerve fiber reconstruction model can be trained using medical images. The nerve fiber reconstruction model can be based on the image information of the medical images and use the prior knowledge about nerve fibers learned by the model during the training phase to reconstruct the image information of the nerve fibers.

[0082] Optionally, the target model can be a pre-trained neural network model, and the training data of the target model includes at least: the original onset area of ​​the brain tumor of the reference subject, the actual recurrence area of ​​the brain tumor, and the image information of the nerve fibers passing through the original onset area and the actual recurrence area at the same time.

[0083] Among them, the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor of the reference subject can be represented in the form of outlined images. The neural network uses the outlined images of these two areas as training data. If a nerve fiber of the reference subject passes through the original onset area and the actual recurrence area at the same time, the nerve fiber is determined to be a high-risk nerve fiber; if a nerve fiber of the reference subject does not pass through the original onset area and the actual recurrence area at the same time, the nerve fiber is determined to be a low-risk nerve fiber. The training goal of the neural network is to accurately judge whether the nerve fiber of the reference subject is a high-risk nerve fiber based on the original onset area of ​​the brain tumor of the reference subject, the actual recurrence area of ​​the brain tumor, and the nerve fiber image information of the reference subject. After the training goal of the neural network is achieved, it can be determined that the target model has been obtained through training.

[0084] As can be seen from the above, the nerve fiber reconstruction unit of the present application uses the target medical image of the target object to simulate and reconstruct the image information of the nerve fibers of the target object, which can provide imaging evidence of the complex interaction between brain tumors and nerve fiber tissues. The visualization of nerve fibers helps to reveal the possible spread paths of brain tumors, which are difficult to effectively determine with traditional imaging observations. Finally, the present application also realizes the transformation from data to knowledge by inputting the target medical image and the reconstructed N nerve fiber image information into a neural network model (target model). Using the target model trained with a large amount of historical data, the key imaging features that affect the risk of recurrence can be automatically extracted, thereby predicting which areas have a higher risk of recurrence, and thus helping doctors to more accurately identify the brain areas that need to be focused on, which is of great reference value for formulating treatment plans such as surgical resection range and radiotherapy target area.

[0085] According to another aspect of an embodiment of the present application, a computer-readable storage medium is further provided, wherein a computer program is stored in the computer-readable storage medium, wherein when the computer program is executed, the device where the computer-readable storage medium is located executes the above-mentioned device for determining the potential recurrence risk area of ​​glioma.

[0086] According to another aspect of an embodiment of the present application, an electronic device is also provided, wherein the electronic device includes one or more processors and a memory, and the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors control the operation of the above-mentioned device for determining the potential recurrence risk area of ​​glioma.

[0087] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0088] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0089] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0090] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0091] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0092] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program code.

[0093] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A device for determining a potential recurrence risk area of ​​a brain tumor, characterized in that: include: a delineation determination unit, configured to determine a brain tumor delineation region of the target object based on a target medical image of the target object; a nerve fiber reconstruction unit, configured to simulate and reconstruct image information of N target nerve fibers of the target object based on the target medical image, wherein N is an integer greater than 1; A risk area prediction unit is used to input the image information of the target medical image and the N target nerve fibers into a target model, perform feature extraction on the received information through the target model, and use the prior knowledge from the model training stage to determine the recurrence risk area of ​​the brain tumor of the target object based on the extracted features.

2. The device for determining a potential recurrence risk area of ​​a brain tumor according to claim 1, wherein: The training data of the target model includes at least: the original onset area of ​​the brain tumor of the reference subject, the actual recurrence area of ​​the brain tumor, and the image information of the nerve fibers passing through the original onset area and the actual recurrence area.

3. The device for determining a potential recurrence risk area of ​​a brain tumor according to claim 1, wherein: The device for determining the potential recurrence risk area of ​​brain tumor also includes: an image processing unit, configured to reconstruct image information of N nerve fibers of a reference object based on a medical image simulation of the reference object; a detection unit, configured to detect, based on image information of the N nerve fibers of the reference subject, whether each nerve fiber of the reference subject simultaneously passes through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor of the reference subject; a high-risk label setting unit, configured to label nerve fibers that simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor as high-risk labels; a low-risk label setting unit, configured to label nerve fibers that do not simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor as low-risk labels; The model training unit is used to iteratively train the neural network based on the image information of the N nerve fibers of the reference object, the risk label marked on each nerve fiber, and the medical image of the reference object to obtain the target model.

4. The device for determining a potential recurrence risk area of ​​a brain tumor according to claim 1, wherein: The delineation determination unit includes: A first delineation subunit is configured to perform image delineation of a brain tumor region based on a medical image of a reference subject taken when the patient first develops the disease, to obtain first delineation information; a second delineation subunit, configured to expand the delineated area in the first delineation information by a preset ratio to obtain second delineation information; The original onset region determination subunit is configured to determine the original onset region of the brain tumor of the reference subject according to the second delineation information.

5. The device for determining a potential recurrence risk area of ​​a brain tumor according to claim 1, wherein: The delineation determination unit includes: a third delineation subunit, configured to perform image delineation of the brain tumor region based on medical images of the reference subject taken when the disease relapsed, to obtain third delineation information; The recurrence region determining subunit is configured to determine the actual recurrence region of the brain tumor of the reference subject according to the third delineation information.

6. The device for determining a potential recurrence risk area of ​​a brain tumor according to claim 1, wherein: The risk area prediction unit includes: a label determination subunit, configured to determine, by using the target model and utilizing the prior knowledge from the model training phase, a risk label carried by each target nerve fiber of the target object according to the extracted features, and to classify target nerve fibers carrying high-risk labels as high-risk nerve fibers; an image reconstruction subunit, configured to reconstruct the high-risk nerve fibers into a 3D image based on a coordinate system corresponding to the target object; an image combination subunit, configured to combine all 3D images obtained based on high-risk nerve fiber reconstruction into a target 3D image; The risk area determination subunit is configured to determine a recurrence risk area of ​​the brain tumor of the target subject based on the target 3D image.

7. The device for determining a potential recurrence risk area of ​​a brain tumor according to claim 6, characterized in that: The risk area determination subunit includes: a detection module, configured to detect the density of high-risk nerve fibers in various areas of the target 3D image; a rendering module, configured to render a region where the density of the high-risk nerve fibers is greater than a preset threshold as a highlighted region; A determination module is configured to determine the highlighted area as a recurrence risk area for the brain tumor of the target subject.

8. The device for determining a potential recurrence risk area of ​​a brain tumor according to claim 5, wherein: The third delineation subunit comprises: a first image processing module, configured to use a medical image of the reference subject taken before the recurrence of the disease as a first image; a second image processing module, configured to use a medical image of the reference subject taken when the disease relapses as a second image; a registration module, configured to register the first image and the second image to obtain a registration result; The delineation module is used to perform image delineation of the brain tumor area according to the registration result to obtain third delineation information.

9. The device for determining a potential recurrence risk area of ​​a brain tumor according to claim 1, wherein: The device for determining the potential recurrence risk area of ​​the brain tumor further comprises: an image acquisition unit, configured to acquire a multimodal medical image of the target object; A data enhancement unit is used to perform data enhancement on the multimodal medical image of the target object and use the multimodal medical image after data enhancement as the target medical image of the target object.

10. A model training method, characterized in that: include: Reconstructing image information of N nerve fibers of the reference object based on medical image simulation of the reference object; determining, based on image information of the N nerve fibers of the reference subject, nerve fibers from the N nerve fibers that simultaneously pass through the original onset area of ​​the brain tumor and the actual recurrence area of ​​the brain tumor of the reference subject; The neural network is iteratively trained based on the image information of the original onset area of ​​the brain tumor of the reference object, the actual recurrence area of ​​the brain tumor, and the nerve fibers that pass through the original onset area and the actual recurrence area at the same time to obtain the target model of any one of claims 1 to 8.

Images