Disease classification model training method and device, equipment and storage medium

Abstract

This application provides a method, apparatus, device, and storage medium for training a disease classification model. The method includes: determining the functional connectivity features of each voxel point in a brain image sample facing a target region; determining key brain regions within the brain image sample based on the functional connectivity features of each voxel point; and training a corresponding disease classification model using the functional connectivity features of each voxel point within the key brain regions. The embodiments of this application can eliminate the interference caused by voxel point features of non-key brain regions in brain image samples on the training of the disease classification model, improve the feature accuracy during disease classification model training, and ensure the training efficiency and classification accuracy of the disease classification model.

Description

Technical Field

[0001] This application relates to the field of image processing technology, specifically to a training method, apparatus, device, and storage medium for a disease classification model. Background Technology

[0002] In the early stages of clinical diagnosis, neuropsychiatric disorders often present with similar clinical manifestations, leading to misdiagnosis. For example, in the clinical diagnosis stage, bipolar disorder (such as mania) and unipolar disorder (such as depression or mania) have similar clinical manifestations and are often misdiagnosed as unipolar disorder.

[0003] Currently, machine learning methods are typically used to train disease classification models by utilizing a large amount of brain imaging data of individuals with a particular brain disease as training samples. However, due to the characteristics of neuropsychiatric disorders, the amount of training samples for a given neuropsychiatric disorder is usually very small, making it impossible to ensure that the disease classification model accurately classifies that neuropsychiatric disorder. Summary of the Invention

[0004] This application provides a training method, apparatus, device, and storage medium for a disease classification model, which eliminates the interference caused by voxel features of non-critical brain regions in brain image samples on the training of the disease classification model, ensuring the training efficiency and classification accuracy of the disease classification model.

[0005] In a first aspect, embodiments of this application provide a method for training a disease classification model, the method comprising:

[0006] Determine the functional connectivity features of each voxel in a brain image sample facing the target region.

[0007] Based on the functional connectivity characteristics of each voxel point, key brain regions within the brain image sample are identified.

[0008] The corresponding disease classification model is trained by utilizing the functional connectivity features of each voxel point in the key brain region.

[0009] Secondly, embodiments of this application provide a training apparatus for a disease classification model, the apparatus comprising:

[0010] The feature determination module is used to determine the functional connectivity features of each voxel in the brain image sample facing the target region.

[0011] A key brain region determination module is used to determine key brain regions within the brain image sample based on the functional connectivity features of each voxel point.

[0012] The model training module is used to train the corresponding disease classification model by utilizing the functional connectivity features of each voxel point in the key brain region.

[0013] Thirdly, embodiments of this application provide an electronic device, which includes:

[0014] A processor and a memory, the memory being used to store a computer program, and the processor being used to call and run the computer program stored in the memory to execute the training method for the disease classification model provided in the first aspect of this application.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium for storing a computer program that causes a computer to execute a training method for a disease classification model as provided in the first aspect of this application.

[0016] Fifthly, embodiments of this application provide a computer program product, including a computer program / instructions, which, when executed by a processor, implements a training method for a disease classification model as provided in the first aspect of this application.

[0017] This application provides a method, apparatus, device, and storage medium for training a disease classification model. First, the functional connectivity features of each voxel point in a brain image sample facing a target region are determined. Then, based on the functional connectivity features of each voxel point, key brain regions within the brain image sample are identified. The functional connectivity features of voxels within these key brain regions are then used to train the corresponding disease classification model. This eliminates interference from voxel features of non-key brain regions within the brain image sample, improving feature accuracy during training and ensuring the training efficiency and classification accuracy of the disease classification model. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a training method for a disease classification model as shown in an embodiment of this application;

[0020] Figure 2 This is a schematic diagram of a brain image sample shown in an embodiment of this application;

[0021] Figure 3 A flowchart illustrating another method for training a disease classification model according to an embodiment of this application;

[0022] Figure 4 This is a schematic block diagram illustrating a training device for a disease classification model, as shown in an embodiment of this application.

[0023] Figure 5 This is a schematic block diagram of an electronic device shown in an embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0025] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0026] To improve the accuracy of disease classification models trained on small sample sets for classifying neuropsychiatric disorders, this application proposes a novel training scheme for disease classification models. By utilizing the functional connectivity features of each voxel point within a brain image sample facing the target region, key brain regions within the brain image sample are identified. Furthermore, the functional connectivity features of each voxel point within the key brain regions are used to train the corresponding disease classification model. This eliminates the interference caused by voxel point features from non-key brain regions within the brain image sample on the training of the disease classification model, improving the feature accuracy during training and ensuring the training efficiency and classification accuracy of the disease classification model.

[0027] Figure 1 This is a flowchart illustrating a training method for a disease classification model, as shown in an embodiment of this application. (Refer to...) Figure 1 The method may include the following steps:

[0028] S110, determine the functional connectivity features of each voxel in the brain image sample facing the target region.

[0029] This application primarily targets neuropsychiatric disorders that originate in the brain and are difficult to identify due to their similar clinical manifestations to other diseases in the early stages of clinical diagnosis. For example, bipolar disorder (such as mania) and unipolar disorder (such as depression or mania) have similar early clinical manifestations and are often misdiagnosed as unipolar disorder.

[0030] To achieve a comprehensive and accurate analysis of any neurological or psychiatric disorder, any scanning device can be used to scan the patient's brain, resulting in corresponding three-dimensional brain images. When training the disease classification model, a large set of three-dimensional brain images from numerous patients can be used as training samples to obtain the brain image samples described in this application.

[0031] As an exemplary solution in this application, the brain image samples in this application are FMRI data obtained after scanning the brains of various patients using functional magnetic resonance imaging (FMRI) technology.

[0032] It should be understood that, because each scan of a patient's brain using fMRI technology typically yields a series of three-dimensional brain images within a very short timeframe, the brain image sample represented by fMRI data is a time series composed of three-dimensional brain images at different points in time.

[0033] In other words, such as Figure 2 As shown, the brain image samples in this application are four-dimensional, containing three-dimensional spatial information and one-dimensional temporal information.

[0034] Furthermore, to ensure the standardization of brain image samples, the following standardization preprocessing operations are typically performed on the brain image samples:

[0035] 1) In order to maintain the magnetization balance within the brain imaging samples, the three-dimensional brain images acquired at the first 10 time points in each brain imaging sample were discarded.

[0036] 2) Perform timed layer correction on the scan layers at all remaining time points within each brain image sample to resolve the layering confusion caused by the interleaved sequences of odd-numbered layers within the brain image sample;

[0037] 3) Time-align the three-dimensional brain images at all time points in each brain image sample with the three-dimensional brain image at the first time point to complete the motion correction and head correction of the brain image sample, so as to eliminate artifacts caused by head movement in the brain image sample.

[0038] 4) Register all data within each brain image sample to the same space to achieve spatial normalization of brain image samples;

[0039] 5) Spatial smoothing is performed on each brain image sample to eliminate noise interference within the brain image sample;

[0040] 6) Regress covariates on the smoothed data of each brain image sample, that is, remove the corresponding interference signals from the time series represented by each brain image sample to reduce the influence of non-neuronal fluctuations in each brain image sample, including white matter signals, cerebrospinal fluid signals and head movement signals.

[0041] 7) Use time bandpass filtering (0.01Hz~0.08Hz) to filter the time series represented by each brain image sample to minimize the impact of low-frequency drift and high-frequency noise within each brain image sample.

[0042] By performing the above data preprocessing on the scanned brain image samples, standardized brain image samples can be obtained to ensure the accuracy of subsequent sample feature extraction.

[0043] In this application, because brain image samples contain three-dimensional spatial information, they can be composed of a large number of voxel points. Furthermore, brain image samples contain a large number of different neurons, and different brain regions have different forms of connections, thus forming a complex and vast brain network. Functional connectivity reflects the systematic organization and interrelationships between different brain regions within the brain space. Methods for measuring and depicting functional connectivity can influence the accuracy of various biomarker identification and individual classification prediction. Therefore, this application can identify corresponding neuropsychiatric disorders by analyzing abnormal functional connectivity relationships between different brain regions. The target region in this application can be a brain region that can elicit different neural responses to emotional fluctuations caused by different neuropsychiatric disorders, such as the bilateral amygdala.

[0044] Furthermore, after acquiring a large number of brain image samples, the functional connectivity features of each voxel can be extracted by analyzing the functional connectivity relationship between each voxel and the target region within the brain image samples.

[0045] As an optional implementation in this application, considering that the locations of functional connectivity abnormalities caused by corresponding emotional fluctuations in the brain due to different neuropsychiatric disorders vary, this application can determine the functional connectivity characteristics of each voxel point within a brain imaging sample through the following steps to ensure the efficiency of neuropsychiatric disorder classification:

[0046] The first step is to identify seed voxels within the target region of the brain imaging sample. Depending on the type of neuropsychiatric disorder, a Region of Interest (ROI) that has a significant impact on the emotional state associated with that disorder can be selected, such as the bilateral amygdala, as the target region in this application. Then, each voxel within this target region is used as a seed voxel for functional connectivity (FC) analysis within the brain imaging sample.

[0047] For example, if the neuropsychiatric disorder is bipolar disorder (such as manic-depressive disorder), then the seed voxel can be the voxel corresponding to the bilateral amygdala regions in the brain image sample.

[0048] The second step is to determine the functional connectivity characteristics of each voxel based on the temporal correlation coefficient between each voxel and the seed voxel in the brain image sample.

[0049] Within the time series represented by a brain image sample, each voxel exhibits corresponding characteristic fluctuations as time points change within the time series. Therefore, this application can determine the characteristic fluctuations of each voxel within the brain image sample over the time series, including the characteristic fluctuations of the seed voxel over the time series.

[0050] Then, by using a corresponding correlation analysis algorithm, such as the Pearson correlation algorithm, the temporal correlation coefficient between each voxel and the seed voxel can be calculated based on the characteristic fluctuations of each voxel in the time series.

[0051] Furthermore, by using the z-transform in the Fisher transform, the temporal correlation coefficient between each voxel and the seed voxel is converted into the corresponding z-score, thereby obtaining the functional connectivity characteristics of each voxel.

[0052] S120 identifies key brain regions within brain image samples based on the functional connectivity characteristics of each voxel.

[0053] Considering that the emotional fluctuations caused by different neuropsychiatric disorders have varying degrees of impact on different brain regions, and consequently, the abnormal effects on functional connectivity within these regions also differ, it is necessary to eliminate voxel features in brain regions that are not closely related to the specific neuropsychiatric disorder from the brain image samples as much as possible to avoid interference with the disease classification model training. This allows for accurate training of the corresponding disease classification model by analyzing voxel features in brain regions that are more relevant to the specific neuropsychiatric disorder.

[0054] Therefore, after determining the functional connectivity features of each voxel, the first step is to obtain an Anatomical Automatic Labeling (AAL) template for brain regions. This AAL template can be obtained by performing standard partitioning of the brain using the AAL algorithm. Then, the brain regions within the brain image sample are determined according to this AAL template.

[0055] It should be understood that the AAL template in this application provides a total of 116 brain regions, but only 90 of them belong to the cerebral structure, while the remaining 26 belong to the cerebellum. Considering that neuropsychiatric disorders mainly involve the cerebral body, this application can focus on analyzing the 90 brain regions within the cerebral structure.

[0056] Furthermore, since different brain regions contain different voxels, the functional connectivity characteristics of each brain region within the brain imaging sample can be preliminarily analyzed based on the functional connectivity characteristics of each voxel. Then, by analyzing the correlation between the functional connectivity characteristics of each brain region and the actual abnormal functional connectivity characteristics of neuropsychiatric diseases, the degree of functional connectivity abnormality of each brain region under the influence of any neuropsychiatric disease can be determined, thereby identifying the brain regions within the brain imaging sample that are more relevant to the neuropsychiatric disease, which will be the key brain regions in this application.

[0057] S130 utilizes the functional connectivity features of voxel points within key brain regions to train corresponding disease classification models.

[0058] After identifying the key brain regions within a brain image sample, the individual voxels contained within those regions can be determined. Then, a preliminary classification model is pre-built, such as a Support Vector Machine (SVM) classifier, linear regression classifier, logistic regression classifier, decision tree classifier, or random forest classifier. The functional connectivity features of each voxel within the key brain region of each brain image sample are continuously input into the pre-built preliminary classification model to train it for classifying any neuropsychiatric disorder until it can accurately classify any neuropsychiatric disorder within any brain image. This yields a trained disease classification model.

[0059] As an optional implementation scheme in this application, considering that key brain regions in brain imaging samples may also contain voxels unrelated to neuropsychiatric diseases, this application utilizes the functional connectivity features of voxels within key brain regions to train the corresponding disease classification model in order to improve the training efficiency of the disease classification model. Specifically, this can be done by: identifying key voxels within key brain regions; and using the functional connectivity features of these key voxels to train the corresponding disease classification model. In other words, some voxels that are more relevant to neuropsychiatric diseases are selected from key brain regions as key voxels, and these are used to train the corresponding disease classification model, thereby improving the training efficiency while ensuring the accuracy of the model training.

[0060] To ensure the accuracy of the disease classification model training, this application divides brain image samples into corresponding training and test sets according to a certain ratio. Then, using each brain image sample in the training set, the corresponding disease classification model is trained following the steps described above. After the disease classification model is trained, the accuracy of the trained model is tested using each brain image sample in the test set to ensure the classification accuracy of the disease classification model for any brain image.

[0061] The technical solution provided in this application first determines the functional connectivity features of each voxel point in a brain image sample facing the target region. Then, based on the functional connectivity features of each voxel point, key brain regions within the brain image sample are identified. The functional connectivity features of voxel points within these key brain regions are then used to train the corresponding disease classification model. This eliminates the interference caused by voxel features from non-key brain regions in the brain image sample on the training of the disease classification model, improves the feature accuracy during training, and ensures the training efficiency and classification accuracy of the disease classification model.

[0062] As an optional implementation of this application, in order to ensure the accuracy of key brain regions in brain image samples, this application can provide a detailed description of the specific steps for determining key brain regions from brain image samples.

[0063] Figure 3 This is a flowchart illustrating another method for training a disease classification model as shown in an embodiment of this application, such as... Figure 3 As shown, the method may include the following steps:

[0064] S310, determine the functional connectivity features of each voxel in the brain image sample facing the target region.

[0065] S320, determine the first importance score of each voxel based on the functional connectivity characteristics of each voxel.

[0066] Given that, to some extent, emotional fluctuations caused by neuropsychiatric disorders often lead to abnormalities in some form of connectivity between related brain regions, the degree of abnormality in the functional connectivity characteristics of different voxels varies under the influence of different neuropsychiatric disorders.

[0067] In this application, by analyzing the correlation between the functional connectivity characteristics of each voxel and the actual abnormal functional connectivity in neuropsychiatric disorders, the first importance score of each voxel can be determined. The first importance score of each voxel can indicate the degree of functional connectivity abnormality of that voxel under the influence of neuropsychiatric disorders.

[0068] As an optional implementation scheme in this application, considering the limited number of brain image samples, to ensure the accuracy of the disease classification model even with a small sample size, this application can employ a corresponding logistic regression algorithm to perform k-fold cross-validation on the functional connectivity features of each voxel, thereby calculating the k importance scores for each voxel. Then, by calculating the average of the k importance scores for each voxel, the first importance score for each voxel can be obtained.

[0069] It should be understood that when determining key voxels within key brain regions, voxels with a first importance score of non-zero within the key brain regions can be directly used as key voxels.

[0070] S330, based on the functional connectivity features and first importance score of each voxel, determines the second importance score and average functional connectivity features of each brain region within the brain image sample.

[0071] Considering that different brain regions within a brain imaging sample contain different voxels, this application can identify the individual voxels contained in each brain region. Then, based on the first importance score of each voxel within each brain region, the second importance score of that brain region is determined. Furthermore, the functional connectivity features of each voxel within each brain region are averaged to obtain the average functional connectivity features of that brain region.

[0072] As an optional implementation of this application, for each brain region within a brain image sample, the sum of the first importance scores of each voxel point within that brain region is calculated as the second importance score of that brain region. That is, the first importance scores of each voxel point contained in each brain region are determined. Then, the sum of the first importance scores of each voxel point within that brain region can be used as the second importance score of that brain region.

[0073] Furthermore, for each brain region within the brain imaging sample, the first importance score of each voxel point within that region is used as the weight of that voxel point. A weighted average of the functional connectivity features of all voxels within that region is then calculated to obtain the average functional connectivity feature of that brain region. In other words, because the first importance scores of each voxel point within each brain region differ, the degree of functional connectivity abnormality of each voxel point under the influence of neuropsychiatric disorders varies. Therefore, to ensure the accuracy of the average functional connectivity feature for each brain region, this application can use the first importance score of each voxel point within any brain region as its weight. Then, using the weight of each voxel point within that brain region, a weighted average of the functional connectivity features of all voxels within that brain region is calculated to obtain the average functional connectivity feature of that brain region.

[0074] S340, based on the correlation coefficient between the average functional connectivity characteristics of each brain region and the pre-defined standard functional connectivity characteristics, and the second importance score of each brain region, key brain regions within the brain imaging sample are identified.

[0075] Considering that the locations of functional connectivity abnormalities caused by emotional fluctuations in the brain vary among different neuropsychiatric disorders, a standard functional connectivity characteristic can be pre-defined for any neuropsychiatric disorder by analyzing the distribution of these abnormalities in the brain when emotional fluctuations cause corresponding functional connectivity abnormalities.

[0076] Then, this application can use a corresponding correlation analysis algorithm to analyze the correlation between the average functional connectivity characteristics of each brain region and the preset standard functional connectivity characteristics, thereby obtaining the correlation coefficient between the average functional connectivity characteristics of each brain region and the standard functional connectivity characteristics. This correlation coefficient can represent the degree of functional connectivity abnormality caused by the emotional fluctuations in the brain caused by mental illness in each brain region.

[0077] Furthermore, since the second importance score of each brain region can also indicate the degree of functional connectivity abnormality of that brain region under the influence of neuropsychiatric diseases, this application, by comprehensively analyzing the correlation coefficient and second importance score corresponding to each brain region, can comprehensively determine the degree of functional connectivity abnormality of each brain region under the influence of neuropsychiatric diseases, thereby identifying the brain regions in the brain imaging sample that are more related to neuropsychiatric diseases as the key brain regions in this application.

[0078] As an optional implementation of this application, the following steps can be used to determine key brain regions within brain image samples:

[0079] The first step is to determine the corresponding brain region boundary score based on the correlation coefficient between the average functional connectivity characteristics of each brain region and the preset standard functional connectivity characteristics, as well as the preset correlation threshold.

[0080] To categorize brain regions within brain imaging samples based on their relevance to the influence of neuropsychiatric disorders, this application pre-sets a correlation threshold according to the specific correlation analysis algorithm employed. After calculating the correlation coefficient between the average functional connectivity characteristics of each brain region and the pre-set standard functional connectivity characteristics using the corresponding correlation analysis algorithm, the correlation coefficient for each brain region can be compared with the correlation threshold one by one, thereby dividing each brain region into two parts: brain regions more relevant to neuropsychiatric disorders and brain regions not relevant to neuropsychiatric disorders. Then, from the second importance scores of the brain regions not relevant to neuropsychiatric disorders, the highest second importance score can be selected, which indicates the brain region most relevant to neuropsychiatric disorders within this part. This highest second importance score is then used as the brain region boundary score in this application.

[0081] Taking the chi-square verification algorithm as an example, when calculating the correlation coefficient between the average functional connectivity feature of each brain region and the pre-defined standard functional connectivity feature, a higher correlation coefficient between a brain region and neuropsychiatric disorders indicates a smaller degree of functional connectivity abnormality in that brain region under the influence of neuropsychiatric disorders, and a lower correlation coefficient indicates a greater degree of functional connectivity abnormality in that brain region under the influence of neuropsychiatric disorders. Therefore, this application sets the correlation threshold for dividing each brain region to 0.05. Assuming the correlation coefficient between the average functional connectivity feature of each brain region and the pre-defined standard functional connectivity feature can be represented as p, then each brain region can be divided into a portion of brain regions unrelated to neuropsychiatric disorders (p>0.05) and another portion of brain regions more related to neuropsychiatric disorders (p<=0.05). Then, in the portion of brain regions unrelated to neuropsychiatric disorders (p>0.05), the highest second importance score is selected as the brain region boundary score in this application.

[0082] The second step is to determine the corresponding key brain regions based on the second importance score and the brain region boundary score for each brain region.

[0083] After determining the brain region boundary score, the second importance score of each brain region can be compared with the brain region boundary score one by one to select the brain regions whose second importance score is greater than the brain region boundary score as key brain regions in the brain imaging sample.

[0084] S350 utilizes the functional connectivity features of voxel points within key brain regions to train corresponding disease classification models.

[0085] The technical solution provided in this application first determines the functional connectivity features of each voxel point in a brain image sample facing the target region. Then, based on the functional connectivity features of each voxel point, key brain regions within the brain image sample are identified. The functional connectivity features of voxel points within these key brain regions are then used to train the corresponding disease classification model. This eliminates the interference caused by voxel features from non-key brain regions in the brain image sample on the training of the disease classification model, improves the feature accuracy during training, and ensures the training efficiency and classification accuracy of the disease classification model.

[0086] Figure 4 This is a schematic diagram illustrating the principle of a training device for a disease classification model, as shown in an embodiment of this application.

[0087] like Figure 4 As shown, the device 400 may include:

[0088] Feature determination module 410 is used to determine the functional connectivity features of each voxel in the brain image sample facing the target region.

[0089] The key brain region determination module 420 is used to determine the key brain regions within the brain image sample based on the functional connectivity features of each voxel point.

[0090] The model training module 430 is used to train the corresponding disease classification model by utilizing the functional connectivity features of each voxel point in the key brain region.

[0091] In some possible implementations, the key brain region identification module 420 may include:

[0092] The first score determination unit is used to determine the first importance score of each voxel point based on the functional connectivity features of each voxel point.

[0093] A brain region feature determination unit is used to determine the second importance score and average functional connectivity feature of each brain region in the brain image sample based on the functional connectivity feature and first importance score of each voxel point.

[0094] The key brain region determination unit is used to determine the key brain regions in the brain image sample based on the correlation coefficient between the average functional connectivity features of each brain region and the preset standard functional connectivity features and the second importance score of each brain region.

[0095] In some feasible implementations, the brain region feature determination unit can be specifically used for:

[0096] For each brain region within the brain image sample, the sum of the first importance scores for each voxel point within that brain region is calculated, and this sum is used as the second importance score for that brain region.

[0097] The first importance score of each voxel point in the brain region is used as the weight of that voxel point. The functional connectivity features of each voxel point in the brain region are weighted and averaged to obtain the average functional connectivity features of the brain region.

[0098] In some feasible implementations, the key brain region identification unit can be specifically used for:

[0099] The corresponding brain region boundary score is determined based on the correlation coefficient between the average functional connectivity feature of each brain region and the preset standard functional connectivity feature and the preset correlation threshold.

[0100] The corresponding key brain regions are determined based on the second importance score and the boundary score of each brain region.

[0101] In some implementations, the model training module 430 can be specifically used for:

[0102] Identify key voxel points within the key brain regions;

[0103] The corresponding disease classification model is trained using the functional connectivity features of the key voxel points.

[0104] In some implementations, the feature determination module 410 can be specifically used for:

[0105] Identify the seed voxel points located within the target region in the brain image sample;

[0106] The functional connectivity features of each voxel are determined based on the temporal correlation coefficient between each voxel in the brain image sample and the seed voxel.

[0107] In this embodiment, the functional connectivity features of each voxel point in the brain image sample facing the target region are first determined. Then, based on the functional connectivity features of each voxel point, key brain regions within the brain image sample are identified. The functional connectivity features of each voxel point within the key brain regions are then used to train the corresponding disease classification model. This eliminates the interference caused by voxel point features of non-key brain regions in the brain image sample on the training of the disease classification model, improves the feature accuracy during disease classification model training, and ensures the training efficiency and classification accuracy of the disease classification model.

[0108] It should be understood that the device embodiments and method embodiments can correspond to each other, and similar descriptions can be referred to the method embodiments. To avoid repetition, further details will not be provided here. Specifically, Figure 4 The apparatus 400 shown can execute any of the method embodiments in this application, and the foregoing and other operations and / or functions of each module in the apparatus 400 are respectively for implementing the corresponding processes in the various methods in the embodiments of this application. For the sake of brevity, they will not be described in detail here.

[0109] The apparatus 400 of this application embodiment has been described above from the perspective of functional modules in conjunction with the accompanying drawings. It should be understood that this functional module can be implemented in hardware, in software instructions, or in a combination of hardware and software modules. Specifically, the steps of the method embodiments in this application can be completed by integrated logic circuits in the processor's hardware and / or by software instructions. The steps of the method disclosed in this application embodiment can be directly embodied as being executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. Optionally, the software module can be located in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps in the above method embodiments.

[0110] Figure 5This is a schematic block diagram of an electronic device shown in an embodiment of this application.

[0111] like Figure 5 As shown, the electronic device 500 may include:

[0112] The system includes a memory 510 and a processor 520. The memory 510 stores computer programs and transfers the program code to the processor 520. In other words, the processor 520 can retrieve and run the computer program from the memory 510 to implement the methods described in the embodiments of this application.

[0113] For example, the processor 520 can be used to execute the above-described method embodiments according to instructions in the computer program.

[0114] In some embodiments of this application, the processor 520 may include, but is not limited to:

[0115] General-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0116] In some embodiments of this application, the memory 510 includes, but is not limited to:

[0117] Volatile memory and / or non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0118] In some embodiments of this application, the computer program may be divided into one or more modules, which are stored in the memory 510 and executed by the processor 520 to perform the method provided in this application. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0119] like Figure 5 As shown, the electronic device may further include:

[0120] Transceiver 530, which can be connected to processor 520 or memory 510.

[0121] The processor 520 can control the transceiver 530 to communicate with other devices; specifically, it can send information or data to other devices or receive information or data sent by other devices. The transceiver 530 may include a transmitter and a receiver. The transceiver 530 may further include antennas, and the number of antennas may be one or more.

[0122] It should be understood that the various components in the electronic device are connected through a bus system, which includes a data bus, a power bus, a control bus, and a status signal bus.

[0123] This application also provides a computer storage medium storing a computer program thereon, which, when executed by a computer, enables the computer to perform the methods of the above-described method embodiments. Alternatively, embodiments of this application also provide a computer program product containing instructions that, when executed by a computer, cause the computer to perform the methods of the above-described method embodiments.

[0124] When implemented using software, it can be implemented entirely or partially as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0125] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0126] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0127] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. For example, the functional modules in the various embodiments of this application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0128] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A training method for a disease classification model, characterized in that, include: The functional connectivity features of each voxel point in a brain image sample facing a target region are determined. The target region is a brain region that makes different neural responses to emotional fluctuations caused by different types of diseases. The functional connectivity features of each voxel point are used to represent the functional connectivity relationship between the voxel point and other voxel points in the target region. Based on the functional connectivity characteristics of each voxel point, key brain regions within the brain image sample are identified. The corresponding disease classification model is trained by utilizing the functional connectivity features of each voxel point in the key brain region. The step of determining key brain regions within the brain image sample based on the functional connectivity features of each voxel point includes: Based on the functional connectivity features of each voxel, determine the first importance score of each voxel. Based on the functional connectivity features and first importance score of each voxel point, the second importance score and average functional connectivity features of each brain region within the brain image sample are determined. Key brain regions within the brain imaging sample are determined based on the correlation coefficient between the average functional connectivity characteristics of each brain region and the preset standard functional connectivity characteristics, as well as the second importance score of each brain region.

2. The method according to claim 1, characterized in that, The step of determining the second importance score and average functional connectivity feature of each brain region within the brain image sample based on the functional connectivity feature and first importance score of each voxel point includes: For each brain region within the brain image sample, the sum of the first importance scores for each voxel point within that brain region is calculated, and this sum is used as the second importance score for that brain region. The first importance score of each voxel point in the brain region is used as the weight of that voxel point. The functional connectivity features of each voxel point in the brain region are weighted and averaged to obtain the average functional connectivity features of the brain region.

3. The method according to claim 1, characterized in that, The process of determining key brain regions within the brain imaging sample based on the correlation coefficient between the average functional connectivity characteristics of each brain region and pre-defined standard functional connectivity characteristics, and the second importance score of each brain region, includes: The corresponding brain region boundary score is determined based on the correlation coefficient between the average functional connectivity feature of each brain region and the preset standard functional connectivity feature and the preset correlation threshold. The corresponding key brain regions are determined based on the second importance score and the boundary score of each brain region.

4. The method according to claim 1, characterized in that, The step of training a corresponding disease classification model using the functional connectivity features of each voxel point within the key brain regions includes: Identify key voxel points within the key brain regions; The corresponding disease classification model is trained using the functional connectivity features of the key voxel points.

5. The method according to claim 1, characterized in that, The determination of the functional connectivity features of each voxel point in the brain image sample facing the target region includes: Identify the seed voxel points located within the target region in the brain image sample; The functional connectivity features of each voxel are determined based on the temporal correlation coefficient between each voxel in the brain image sample and the seed voxel.

6. A training device for a disease classification model, characterized in that, include: The feature determination module is used to determine the functional connectivity features of each voxel point in the brain image sample facing the target region. The target region is the brain region that makes different neural responses to emotional fluctuations caused by different types of diseases. The functional connectivity features of each voxel point are used to represent the functional connectivity relationship between the voxel point and other voxel points in the target region. A key brain region determination module is used to determine key brain regions within the brain image sample based on the functional connectivity features of each voxel point. The model training module is used to train the corresponding disease classification model by utilizing the functional connectivity features of each voxel point in the key brain region. Specifically, the key brain region identification module is used for: Based on the functional connectivity features of each voxel, determine the first importance score of each voxel. Based on the functional connectivity features and first importance score of each voxel point, the second importance score and average functional connectivity features of each brain region within the brain image sample are determined. Key brain regions within the brain imaging sample are determined based on the correlation coefficient between the average functional connectivity characteristics of each brain region and the preset standard functional connectivity characteristics, as well as the second importance score of each brain region.

7. An electronic device, characterized in that, include: A processor and a memory, the memory being used to store a computer program, the processor being used to call and run the computer program stored in the memory to perform the training method of the disease classification model according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, Used to store computer programs that cause a computer to perform a training method for a disease classification model as described in any one of claims 1-5.

9. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the training method for the disease classification model as described in any one of claims 1-5.

Images