Hippocampus abnormal atrophy identification processing method of craniocerebral function nuclear magnetic image

Abstract

The invention discloses a hippocampus abnormal atrophy identification processing method of a craniocerebral function nuclear magnetic resonance image. The method comprises the following steps: acquiring to-be-processed craniocerebral function nuclear magnetic resonance image data; preprocessing the craniocerebral function nuclear magnetic resonance image data to obtain preprocessed craniocerebral function nuclear magnetic resonance image data; extracting a hippocampus area in the preprocessed craniocerebral functional nuclear magnetic resonance image data; establishing a hippocampus three-dimensional image based on the hippocampus area; performing feature extraction on the hippocampus three-dimensional image of the hippocampus region based on a deep learning network to obtain a feature extraction result; and according to the feature extraction result, carrying out abnormal atrophy classification identification on the hippocampus body region to obtain a classification identification result.

Description

Technical Field

[0001] The present invention relates to the field of image recognition, and in particular to a method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging image. Background Art

[0002] Alzheimer's disease (AD) is a neurodegenerative disease with progressive cognitive dysfunction and behavioral abnormalities as its main clinical manifestations. With the accelerated aging of the global population, the incidence of AD has shown a significant upward trend. Studies have shown that AD accounts for 50%-70% of all cases of dementia, and has become a major public health issue threatening the health of the elderly.

[0003] There are currently about 50 million dementia patients worldwide, and this number is expected to increase to 152 million by 2050. Age is the main risk factor for AD, with about 8% of people over 65 suffering from the disease, and the prevalence rate for people over 85 years old is as high as 30%. There are about 10 million AD patients in my country, and the number is increasing by 300,000 per year. Early diagnosis and intervention are of great significance to improving the prognosis of patients.

[0004] At present, traditional detection methods for AD mainly rely on clinical symptom assessment, cerebrospinal fluid biomarker detection and imaging examinations; however, these methods can often only make a clear diagnosis when the disease develops to the middle and late stages, which seriously affects the effectiveness of early intervention. Hippocampal atrophy is one of the earliest and most significant imaging changes in AD. Studies have shown that the hippocampus of AD patients shrinks 2-3 times faster than that of normal people. Therefore, accurately identifying and quantifying the degree of hippocampal atrophy is of great significance for the early diagnosis of AD.

[0005] Traditional diagnostic methods are mainly based on image detection and recognition methods of traditional imaging, namely computer-aided diagnosis systems, and their application in early diagnosis of AD has become a research hotspot. They mainly include three categories based on traditional image processing, machine learning and deep learning. In image processing, threshold segmentation, region growing, edge detection and other methods are mainly used to extract the hippocampus region, but these methods have high requirements on image quality and are easily affected by noise. In feature extraction, they mainly include morphological features (volume, surface area, etc.), statistical features (grayscale histogram, co-occurrence matrix, etc.) and texture features (wavelet coefficients, fractal dimension, etc.). Although these features can reflect some characteristics of the hippocampus, they often require manual design and selection, and it is difficult to fully describe complex pathological changes. In terms of classification and recognition, classic algorithms such as support vector machine (SVM), random forest (RF), and K nearest neighbor (KNN) are commonly used, but these methods are highly dependent on the expression ability of features and are difficult to process high-dimensional data. In recent years, some studies have begun to try to introduce deep learning methods, but due to the difficulty in obtaining medical imaging data, high annotation costs, and generally insufficient sample size, the performance of deep learning models is limited. In addition, most existing studies use unified evaluation criteria and fail to fully consider the impact of age on hippocampal atrophy, which can easily lead to diagnostic bias.

[0006] The main challenge faced by traditional image detection and recognition methods in hippocampal atrophy recognition is the limitation of feature extraction. First, manually designed features are often based on expert experience, and it is difficult to ensure their integrity and representativeness. For example, simple volume measurements cannot reflect the fine structural changes of the hippocampus, and complex texture features are easily affected by image quality. Secondly, the feature selection process lacks theoretical guidance, and trial and error or statistical significance tests are often used, which makes it difficult to ensure that the selected features have the best discriminative ability. Thirdly, there may be correlations between different features, resulting in information redundancy and increased computational complexity. In addition, traditional feature extraction methods are difficult to adapt to image differences caused by different scanning devices and parameters, and their generalization ability is limited. In terms of feature fusion, simple feature splicing or weighted combination is difficult to fully utilize the complementarity between different features, affecting the final classification effect. These limitations make it difficult for traditional machine learning methods to achieve ideal diagnostic accuracy in practical applications.

[0007] In addition, there are obvious inconsistencies in the current image detection and recognition methods for hippocampal atrophy, especially the lack of stratification standards that take age factors into account. Studies have shown that the hippocampus also experiences a volume reduction during normal aging, but its atrophy rate and pattern are different from pathological atrophy. Existing evaluation methods often use a unified threshold standard, which ignores the effect of age on hippocampal volume and may lead to an increase in false positive rates in the elderly and an increase in false negative rates in the young. In addition, the measurement methods and scoring criteria used in different studies are not the same and lack comparability. This inconsistency in evaluation standards not only affects the accuracy of diagnosis, but also brings difficulties to the promotion and application of clinical practice and research results. Therefore, establishing a standardized evaluation system based on age stratification has become an urgent problem to be solved. Summary of the invention

[0008] The purpose of the present invention is to provide a method for identifying and processing abnormal atrophy of the hippocampus in functional magnetic resonance imaging of the brain, which solves the above-mentioned technical problems pointed out in the prior art.

[0009] The present invention provides a method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging, comprising the following steps:

[0010] Acquiring brain functional magnetic resonance imaging data to be processed;

[0011] Preprocessing the brain functional magnetic resonance image data to obtain preprocessed brain functional magnetic resonance image data;

[0012] Extracting the hippocampus region from the preprocessed brain functional magnetic resonance image data; and establishing a three-dimensional image of the hippocampus based on the hippocampus region;

[0013] Performing feature extraction on the three-dimensional image of the hippocampus region based on a deep learning network to obtain a feature extraction result;

[0014] According to the feature extraction result, the abnormal atrophy classification and identification of the hippocampus region is performed to obtain a classification and identification result.

[0015] Preferably, after obtaining the brain functional magnetic resonance imaging data to be processed, the following operation steps are also included:

[0016] Obtaining age information of a subject corresponding to the brain functional magnetic resonance imaging data to be processed;

[0017] Extracting hippocampus region data in the coronal T1-weighted image in the hippocampus three-dimensional image;

[0018] Establishing a quantitative assessment model for the degree of hippocampal atrophy based on the hippocampal region data;

[0019] The age information A, the quantitative evaluation model H for the degree of hippocampal atrophy and the feature extraction result F are combined to calculate the grading of hippocampal atrophy and obtain a grading result S.

[0020] Preferably, the hippocampus region data includes hippocampus volume data and morphological characteristic parameter data;

[0021] The quantitative evaluation model for the degree of hippocampal atrophy includes a hippocampal atrophy rate calculation model, a hippocampal volume loss speed calculation model, and a normal reference value database based on age stratification;

[0022] Preferably, the calculation of hippocampal atrophy grading determination by combining the age information A, the hippocampal atrophy degree quantitative assessment model H and the feature extraction result F to obtain the grading determination result S includes the following steps:

[0023] Based on the age information A, the normal reference value data corresponding to the age information A is retrieved from the normal reference value database based on age stratification;

[0024] A quantitative evaluation model H' for the degree of hippocampal atrophy after retrieval is established based on the normal reference value data corresponding to the age information A, the hippocampal atrophy rate calculation model, and the hippocampal volume loss speed calculation model;

[0025] The age information A, the quantitative evaluation model H' for the degree of hippocampal atrophy after retrieval, and the feature extraction result F are standardized to obtain standardized age information A', the standardized quantitative evaluation model H' for the degree of hippocampal atrophy after retrieval, and the standardized feature extraction result F';

[0026] Calculating the graded judgment result S based on the standardized age information A', the standardized post-retrieval hippocampal atrophy degree quantitative assessment model H", and the standardized feature extraction result F', to obtain the graded judgment result S;

[0027] The calculation method of the classification judgment result S is:

[0028] S=w1×A′+w2×H”+w3×F′+b;

[0029] Wherein, w1, w2 and w3 are weight coefficients respectively; b is the bias term; A' is the standardized age information; H" is the standardized quantitative assessment model of hippocampal atrophy; and F' is the standardized feature extraction result.

[0030] Preferably, the preprocessing of the brain functional magnetic resonance image data to obtain the preprocessed brain functional magnetic resonance image data comprises the following steps:

[0031] The non-local mean filtering algorithm is used to denoise the brain functional magnetic resonance image data to obtain denoised brain functional magnetic resonance image data;

[0032] Performing grayscale value standardization on the denoised brain functional magnetic resonance image data to obtain grayscale standardized brain functional magnetic resonance image data;

[0033] The grayscale-normalized functional magnetic resonance imaging data of the brain are spatially registered with a preset standard template using an affine transformation method to obtain a registered image.

[0034] The registered images are corrected for deviation fields to obtain preprocessed brain functional magnetic resonance image data.

[0035] Preferably, the deviation field correction is performed on the registered image to obtain the pre-processed brain functional magnetic resonance image data, comprising the following steps:

[0036] Performing Fourier transform processing on the registered image to obtain a low-frequency image;

[0037] Based on the low-frequency image as an initial estimate, an iterative optimization processing operation is performed through the N4ITK algorithm to obtain pre-processed brain functional magnetic resonance imaging data.

[0038] Preferably, the iterative optimization processing operation is performed by using the N4ITK algorithm based on the low-frequency image as an initial estimate to obtain the pre-processed brain functional magnetic resonance image data, including the following operating steps:

[0039] Initialize iteration parameters;

[0040] The iteration parameters include a low-frequency image L(i,j), a deviation field matrix B(i,j), a B-spline grid spacing d, a regularization weight coefficient λ, a spline basis function matrix W(i,j), a learning rate α, and an energy numerical function E(i,j);

[0041] Calculating based on the low-frequency image L(i,j) and the deviation field matrix B(i,j) to obtain a first corrected image C(i,j);

[0042] Based on a search window of a preset size, the first corrected image to be determined is searched to obtain a local area image matrix; based on the local area image matrix and the energy value function, a local energy value E is calculated. local(i,j) ;

[0043] Determine the local energy value E local(i,j) Is it greater than the preset local energy value minimum threshold P? If so, the local energy value E is adjusted based on the learning rate α. local(i,j)The corresponding local area image matrix is ​​locally iteratively optimized to obtain an optimized local area image matrix;

[0044] Constructing a second corrected image to be determined based on the optimized local image matrix;

[0045] Performing deviation field estimation based on the second corrected image to be determined and the low-frequency image L(i,j) to obtain an updated deviation field matrix B'(i,j);

[0046] Based on the second corrected image to be determined and the updated deviation field matrix B'(i,j), a corrected energy value E'(i,j) is calculated by using an energy value function E(i,j);

[0047] Determine whether the corrected energy value E'(i,j) is greater than or equal to the preset minimum threshold of the total energy function; if so, output the second corrected image to be determined as the preprocessed brain functional magnetic resonance image data; if not, return to the above step S2422 and iterate again until the preprocessed brain functional magnetic resonance image data is output.

[0048] Preferably, the energy value function E(i,j) is calculated as follows:

[0049]

[0050] Where N and M are the number of spline basis functions and bias fields used for calculation, respectively; L(i,j) is the low-frequency image; B(i,j) is the bias field matrix; d is the B-spline grid spacing; λ is the regularization weight coefficient; W(i,j) is the spline basis function matrix.

[0051] Preferably, the local energy value E is calculated based on the learning rate α. local(i,j) The corresponding local area image matrix is ​​locally iteratively optimized to obtain an optimized local area image matrix, including the following steps:

[0052] Based on the learning rate, the local area image matrix Q is calculated by using the gradient descent method. local(i,j) Perform local optimization to obtain a first optimized image matrix;

[0053] The energy of the first optimized image matrix is ​​calculated based on the energy value function E(i, j) to obtain a first local energy value E' local(i,j) ;

[0054] Determine the first local energy value E' local(i,j)Is it greater than the preset local energy value minimum threshold value P? If not, then output the first optimized image matrix as the optimized local area image matrix; if so, then further determine the first local energy value E' local(i,j) whether the energy ratio to the local energy value minimum threshold value P is greater than or equal to the preset step size adjustment threshold; if so, adjusting the learning rate α based on the energy ratio to obtain an adjusted first learning rate α';

[0055] The first learning rate α' and the first optimized image matrix are returned to the above operation for re-iteration until an optimized local area image matrix is ​​output.

[0056] Preferably, the first optimized image matrix is ​​calculated as follows:

[0057]

[0058] In the formula, Represents the local energy value E local(i,j) Partial derivatives of the pixel values ​​of each pixel point of the image matrix in the current local area;

[0059] The first learning rate α' is calculated as follows:

[0060]

[0061] Among them, β is the step size adjustment coefficient.

[0062] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages:

[0063] By analyzing the above-mentioned method for identifying and processing abnormal atrophy of the hippocampus in the functional magnetic resonance imaging of the brain provided by the present invention, it can be known that in specific applications, the functional magnetic resonance imaging data of the brain of the patient, which contains the functional activities, structure and potential pathological information of the brain, is collected by medical imaging equipment (such as magnetic resonance imaging equipment) to provide original input for subsequent analysis; further, by performing preprocessing operations such as deviation field correction processing on the collected image data, high-quality input is provided for subsequent feature extraction and modeling; then, the hippocampus region is extracted from the preprocessed image data, and a three-dimensional image of the hippocampus is created based on the region, and the hippocampus region is accurately located and extracted, and then three-dimensional reconstruction is performed using this information. It can present the morphological characteristics of the hippocampus in more detail, especially when evaluating changes in the volume of the hippocampus (such as atrophy). Furthermore, it can use deep learning networks (such as multi-scale convolutional neural networks) to extract features from three-dimensional hippocampal images to obtain the shape, volume and texture features of the hippocampus. The combination of multi-scale convolutional neural networks and attention mechanisms can improve the model's adaptability to changes under different image conditions, improve the accuracy and comprehensiveness of feature extraction, and more accurately reflect the pathological characteristics of the hippocampus. Finally, based on the extracted features, the abnormal atrophy of the hippocampus is classified and identified, and classification results (such as normal or abnormal) are generated, which provides an effective decision support tool for clinicians to assist doctors in formulating treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A schematic diagram of the main process of a method for identifying and processing abnormal atrophy of the hippocampus in functional magnetic resonance imaging of the brain;

[0065] Figure 2 A schematic diagram of a three-dimensional hippocampal region image in a method for identifying and processing abnormal hippocampal atrophy in a functional magnetic resonance imaging of the brain;

[0066] Figure 3 A schematic diagram of pons imaging in a method for identifying and processing abnormal hippocampal atrophy in functional brain magnetic resonance imaging;

[0067] Figure 4 A schematic diagram of a coronal image in a method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging;

[0068] Figure 5 The present invention is a schematic diagram of the operation flow of the pre-processed functional magnetic resonance imaging data of the brain obtained by iterative optimization processing in a method for identifying abnormal atrophy of the hippocampus in functional magnetic resonance imaging of the brain. DETAILED DESCRIPTION

[0069] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] The present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings.

[0071] like Figure 1 As shown, the embodiment of the present invention provides a method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging, comprising the following steps:

[0072] Step S10: Acquire brain functional magnetic resonance imaging data to be processed;

[0073] It should be noted that the above-mentioned brain functional MRI image data to be processed are the patient's brain functional MRI image data to be processed acquired by medical imaging equipment (such as magnetic resonance imaging equipment), which can capture the functional activities, structures and possible lesions of the brain. The above-mentioned brain functional MRI image data to be processed are multiple planar slices in two-dimensional form (such as cross-sections or sagittal planes, etc.), and each slice represents a different brain cross-section.

[0074] Step S20: preprocessing the brain functional magnetic resonance image data to obtain preprocessed brain functional magnetic resonance image data;

[0075] The preprocessing includes bias field correction processing;

[0076] It should be noted that in the above-mentioned embodiments of the present application, a low-frequency, gradual intensity inhomogeneity, called a bias field (BiasField), is usually generated in the functional magnetic resonance imaging data of the brain due to reasons such as magnetic field inhomogeneity and detector response differences; the bias field is usually manifested as uneven brightness of the image, which will affect the quantitative analysis of the image and subsequent processing tasks. Therefore, bias field correction is an important step to improve image quality and subsequent analysis accuracy; the main goal of bias field correction is to remove or reduce this low-frequency, uneven brightness change from the image, so that the image can more accurately reflect the true physical characteristics of the tissue, especially in subsequent image analysis tasks (such as tissue segmentation, quantitative analysis, etc.), removing the bias field can significantly improve the accuracy of the analysis results.

[0077] Step S30: extracting the hippocampus region in the preprocessed brain functional magnetic resonance image data; and establishing a three-dimensional image of the hippocampus based on the hippocampus region;

[0078] It should be noted that the above-mentioned three-dimensional hippocampus image refers to a three-dimensional hippocampus region image constructed based on the hippocampus region image in the pre-processed brain functional magnetic resonance image data (i.e., a three-dimensional image, or a three-dimensional reconstructed image, such as Figure 2 As shown); The above-mentioned embodiment of the present application can accurately extract the hippocampus area and perform three-dimensional modeling, so as to perform a more detailed analysis of the hippocampus, and can more accurately evaluate its morphological changes (such as atrophy or hyperplasia), which is helpful for disease diagnosis; Under normal circumstances, the functional magnetic resonance imaging data of the brain can be in two-dimensional form or three-dimensional form. However, the two-dimensional form (i.e., the functional magnetic resonance imaging data of the brain to be processed) is presented in the form of a two-dimensional image, which is not strong for complex structures and it is difficult to predict the volume and atrophy of the hippocampus; The three-dimensional form is a volume data set collected by a magnetic resonance imaging device, which can reconstruct continuous two-dimensional slices in any direction or directly display them in the form of a three-dimensional model. This type of imaging provides more detailed information, especially for complex structures such as the brain, heart or joints, etc., and can provide better anatomical details. However, 3DMRI may require more scanning time and has higher requirements for equipment;

[0079] Based on this, the embodiment of the present application acquires the to-be-processed cranial functional MRI image data in two-dimensional form, then extracts the hippocampus region from the to-be-processed cranial functional MRI image data in two-dimensional form, and then performs three-dimensional reconstruction on the hippocampus region corresponding to multiple plane slices in a targeted manner, thereby obtaining a three-dimensional image of the hippocampus region in three-dimensional form, thereby reducing the scanning time of the MRI device while displaying the loaded brain structure.

[0080] Step S40: extracting features of the three-dimensional image of the hippocampus region based on a deep learning network to obtain a feature extraction result;

[0081] It should be noted that the features extracted by the deep learning network in the above-mentioned embodiments of the present application can usually capture potential patterns or subtle changes that are difficult to detect with traditional methods;

[0082] Specifically, the feature extraction includes the following steps: first, construct a multi-scale convolutional neural network for feature learning to capture feature information of different scales so that the network can effectively process objects of different sizes and shapes, enhance the network's robustness to image deformation and scale changes, enable it to stably perform under diverse image conditions in the real world, improve feature expression capabilities, fuse features at different levels, enhance the ability to recognize complex patterns, reduce information loss, retain important global and local information, adapt to diverse inputs, and improve the generalization ability of the model; then, extract the shape features, volume features and texture features of the hippocampus through the feature learning results; further, use the attention mechanism to weightedly fuse the shape features, volume features and texture features of the horse's body to obtain fused features; finally, use the dimensionality reduction algorithm to optimize the fused features to obtain feature extraction results.

[0083] Step S50: Based on the feature extraction result, the abnormal atrophy of the hippocampus region is classified and identified to obtain a classification identification result.

[0084] It should be noted that the above-mentioned embodiments of the present application can quantitatively evaluate the health status of the hippocampus through classification and identification, and provide reliable decision support for clinicians. At the same time, the classification method based on deep learning usually has high accuracy and reliability, and can effectively distinguish between normal and abnormal states.

[0085] It should be noted that the above-mentioned embodiment of the present application collects the patient's cranial functional magnetic resonance imaging data containing the functional activities, structure and potential pathological information of the brain through medical imaging equipment (such as magnetic resonance imaging equipment), so as to provide original input for subsequent analysis; further, by performing preprocessing operations such as deviation field correction processing on the collected image data, high-quality input is provided for subsequent feature extraction and modeling; then, the hippocampus region is extracted from the preprocessed image data, and a three-dimensional image of the hippocampus is created based on the region, and the hippocampus region is accurately located and extracted, and then three-dimensional reconstruction is performed using this information to present the morphological characteristics of the hippocampus in more detail. It is particularly advantageous when evaluating changes in the volume of the hippocampus (such as atrophy). Furthermore, a deep learning network (such as a multi-scale convolutional neural network) is used to extract features from three-dimensional hippocampal images to obtain the shape, volume, and texture features of the hippocampus. The combination of a multi-scale convolutional neural network and an attention mechanism improves the model's ability to adapt to changes under different image conditions, improves the accuracy and comprehensiveness of feature extraction, and more accurately reflects the pathological characteristics of the hippocampus. Finally, based on the extracted features, the abnormal atrophy of the hippocampus is classified and identified, and a classification result (such as normal or abnormal) is generated, which provides an effective decision support tool for clinicians and assists doctors in formulating treatment plans.

[0086] Specifically, in step S10, after obtaining the brain functional magnetic resonance image data to be processed, the following operation steps are also included:

[0087] Step S60: obtaining age information of the subject corresponding to the brain functional magnetic resonance image data to be processed;

[0088] Step S70: extracting hippocampus region data in the coronal T1-weighted image in the hippocampus three-dimensional image;

[0089] The hippocampus region data includes hippocampus volume data and morphological feature parameter data;

[0090] It should be noted that the above extraction of the hippocampus region first locates the horizontal layer of the anterior pons, such as Figure 3 As shown, the pons is a structure located in the brainstem, between the medulla oblongata and the midbrain. Its position is relatively fixed in brain images and can usually be used as a reference for positioning. The horizontal plane of the anterior pons can be located by manually marking or setting a reference plane based on anatomical knowledge.

[0091] Furthermore, after locating the horizontal plane of the anterior pons, in the coronal T1-weighted image, selecting the plane passing through the hippocampus can accurately extract the structural information of the hippocampus. The hippocampus region is usually located on the medial side of the temporal lobe, and its main function involves memory storage and processing. The T1-weighted image has good tissue contrast, especially in the imaging of brain structures such as the hippocampus. Coronal plane refers to the slicing method from front to back, such as Figure 4 As shown, the morphology of the hippocampus can usually be displayed more accurately on this plane; by analyzing the coronal image sequence, the slice planes related to the hippocampus can be identified, and these planes are usually located in the medial temporal lobe, through the cross section of the hippocampus;

[0092] Furthermore, the hippocampus contour is obtained through a pre-trained deep learning segmentation network (such as a convolutional neural network (CNN); then, the hippocampus volume and morphological feature parameters are calculated. Specifically, the hippocampus volume is estimated by calculating the number of voxels contained in the segmentation contour of the hippocampus region, and the surface area of ​​the hippocampus is calculated by performing surface reconstruction on the segmentation contour of the hippocampus (such as using the MarchingCubes algorithm). The geometric features of the hippocampus, such as symmetry, shape factor, curvature, etc., are extracted using a morphological feature extraction algorithm (such as region-based features, boundary features, etc.). By performing statistical analysis on the grayscale values ​​in the hippocampus region, some features related to atrophy (such as gray matter density, texture features, etc.) can be extracted.

[0093] Step S80: establishing a quantitative assessment model for the degree of hippocampal atrophy based on the hippocampal region data;

[0094] The quantitative evaluation model for the degree of hippocampal atrophy includes a hippocampal atrophy rate calculation model, a hippocampal volume loss speed calculation model, and a normal reference value database based on age stratification;

[0095] Step S90: Calculate the hippocampal atrophy grading based on the age information A, the quantitative assessment model H for the degree of hippocampal atrophy (definitely incorrect), and the feature extraction result F to obtain a grading result S (the grading result is the classification result in the above step S50).

[0096] It should be noted that the calculation result of the above-mentioned grading determination result S is a numerical value. After the above-mentioned grading determination result is calculated, it is also necessary to judge the numerical value of the grading determination result with the preset grading numerical range. When the grading determination result is less than or equal to the mild atrophy threshold S1, the grading determination result is a mild atrophy level; when the grading determination result is greater than the mild atrophy threshold S1 and less than or equal to the moderate atrophy threshold S2, the grading determination result is moderate atrophy; when the grading determination result is greater than the moderate atrophy threshold S2, the grading determination result is severe atrophy;

[0097] The above-mentioned embodiment of the present application first obtains the age information of the subject corresponding to the brain functional magnetic resonance imaging data to be processed. Age is one of the important factors of hippocampal atrophy. Usually, the hippocampus may atrophy with the increase of age, so age information is crucial for the subsequent atrophy assessment; by introducing age information, a physiological background is provided for the evaluation of hippocampal atrophy, because normal hippocampal atrophy is closely related to age, especially in the elderly; combined with age information, it can be compared with normal reference values ​​based on age stratification to determine whether the subject's hippocampus has abnormal atrophy, avoiding misdiagnosis or misevaluation caused by ignoring age differences in traditional hippocampal atrophy analysis and processing methods; further, the hippocampal regional data in the coronal T1-weighted image is extracted from the three-dimensional image of the hippocampus; the coronal image has good structural contrast, can accurately display the morphology of the hippocampus, and provide a clear basic image for subsequent analysis; T The weighted image can effectively distinguish brain structures, and is particularly suitable for displaying the structure of the hippocampus. Its high contrast makes the details of the hippocampus in the image more obvious, and can accurately capture atrophy or other morphological changes. Further, the extracted hippocampal regional data (volume and morphological features) are applied to the quantitative evaluation model of the hippocampal atrophy degree to achieve quantitative analysis of the hippocampal atrophy. It is particularly important to note that the reference database based on age stratification is particularly important because the atrophy of the hippocampus is not only related to diseases (such as Alzheimer's disease), but also to the normal aging process. By comparing with normal reference values, physiological atrophy and pathological atrophy can be effectively distinguished. Finally, the age information obtained in step S60, the quantitative evaluation model of the hippocampal atrophy degree (H) established in step S80, and the feature extraction result (F) extracted in step S40 are combined to perform a hippocampal atrophy grading judgment, and the grading judgment result (S) is output, that is, whether there is abnormal atrophy of the hippocampus.

[0098] Specifically, in step S90, the hippocampal atrophy grading determination is calculated by combining the age information A, the hippocampal atrophy degree quantitative assessment model H and the feature extraction result F to obtain the grading determination result S, including the following operation steps:

[0099] Step S91: based on the age information A, retrieve the normal reference value data corresponding to the age information A from the normal reference value database based on age stratification;

[0100] Step S92: A quantitative evaluation model H' of the degree of hippocampal atrophy after retrieval is established based on the normal reference value data corresponding to the age information A, the hippocampal atrophy rate calculation model, and the hippocampal volume loss speed calculation model;

[0101] Step S93: Standardizing the age information A, the quantitative evaluation model H' for the degree of hippocampal atrophy after retrieval, and the feature extraction result F to obtain standardized age information A', the standardized quantitative evaluation model H' for the degree of hippocampal atrophy after retrieval, and the standardized feature extraction result F';

[0102] Step S92: Calculate the graded judgment result S based on the standardized age information A', the standardized post-retrieval hippocampal atrophy degree quantitative assessment model H", and the standardized feature extraction result F' to obtain the graded judgment result S;

[0103] The calculation method of the classification judgment result S is:

[0104] S=w1×A′+w2×H”+w3×F′+b;

[0105] Wherein, w1, w2 and w3 are weight coefficients respectively; b is the bias term; A' is the standardized age information; H" is the standardized quantitative evaluation model of hippocampal atrophy; F' is the standardized feature extraction result;

[0106] It should be noted that the above-mentioned embodiment of the present application standardizes the age information A, the quantitative assessment model H for the degree of hippocampal atrophy, and the feature extraction result F to ensure that features of different scales can be compared and calculated in the same way, thereby more accurately assessing the degree of hippocampal atrophy.

[0107] Specifically, in step S80, a quantitative evaluation model for the degree of hippocampal atrophy is established based on the hippocampal region data, including the following steps:

[0108] Step S81: constructing a hippocampal atrophy rate calculation model based on the hippocampal volume data of multiple time nodes; and obtaining a hippocampal volume loss rate calculation model based on the initial hippocampal volume data of the initial time node and the final hippocampal volume data of the final time node among the multiple time nodes and the time interval between the initial time node and the final time node;

[0109] The calculation method of the hippocampus annual atrophy rate calculation model is:

[0110]

[0111] The calculation method of the hippocampus volume loss rate calculation model is:

[0112]

[0113] It should be noted that the above-mentioned embodiment of the present application is to collect hippocampal volume data at multiple time nodes (for example, every few years or every few months); then, the annual atrophy rate of the hippocampal volume data at each time node corresponding to the hippocampal volume data at the previous time node is calculated based on the hippocampal volume data at multiple time nodes; the atrophy rate of the hippocampus is evaluated by the above-mentioned hippocampal atrophy rate calculation model; generally, a higher annual atrophy rate usually indicates more significant atrophy, which may be related to certain neurodegenerative diseases (such as Alzheimer's disease), and serves as a reference for early disease monitoring.

[0114] Step S82: Establishing a normal reference value database based on age stratification by counting the hippocampal volume data corresponding to multiple age groups of multiple normal subjects;

[0115] It should be noted that the above-mentioned embodiments of the present application collect hippocampal volume data at multiple time nodes to establish a hippocampal atrophy rate calculation model and a hippocampal volume loss speed calculation model, thereby quantitatively evaluating the atrophy rate of the hippocampus; by introducing multi-time node data, the changes in the hippocampus can be dynamically tracked. Compared with static analysis (data at a single time point), analysis of multiple time nodes can more accurately evaluate the atrophy trend of the hippocampus and provide a basis for early detection and monitoring of the disease; the annual atrophy rate is an important indicator that can help clinicians understand the progression rate of hippocampal atrophy. A higher annual atrophy rate usually means significant atrophy, which may be a neurodegenerative disease. The model is a clinical manifestation of dementia (such as Alzheimer's disease). Through long-term dynamic data monitoring, it can provide data support for the early diagnosis of the disease. The hippocampal volume loss rate calculation model provides a more detailed assessment of the atrophy rate. By comparing the volume changes between different time nodes, it can reveal the rate of atrophy and then determine whether there is abnormal rapid atrophy, which is critical for early intervention and treatment. Furthermore, by statistically analyzing the hippocampal volume data of multiple normal subjects at different ages, a normal reference value database based on age stratification is established as a standard reference for subsequent hippocampal atrophy assessments, helping to determine whether there are abnormalities in the subject's hippocampus.

[0116] Specifically, in step S20, the brain functional magnetic resonance image data is preprocessed to obtain preprocessed brain functional magnetic resonance image data, including the following operation steps:

[0117] Step S21: using a non-local mean filtering algorithm to perform denoising on the brain functional magnetic resonance image data to obtain denoised brain functional magnetic resonance image data;

[0118] Step S22: performing grayscale value standardization processing on the denoised brain functional magnetic resonance image data to obtain grayscale standardized brain functional magnetic resonance image data;

[0119] Step S23: using an affine transformation method to spatially register the grayscale-normalized brain functional magnetic resonance image data with a preset standard template to obtain a registered image;

[0120] Step S24: performing deviation field correction on the registered image to obtain pre-processed brain functional magnetic resonance image data.

[0121] It should be noted that the above-mentioned embodiment of the present application first uses a non-local mean filtering algorithm to denoise the brain functional magnetic resonance image data. Denoising is an important step in MRI image processing, because brain functional magnetic resonance images are often affected by various noises (such as thermal noise, motion artifacts, etc.), which affect the quality of the image. The denoised image can more accurately reflect the structure and function of the brain, ensure the reliability of subsequent analysis, and can better balance noise removal and detail retention. It is suitable for situations where fine structure retention is required in medical imaging, such as the evaluation of small structures such as the hippocampus; further, the grayscale range of the image is unified through grayscale value standardization processing, so that the data of different images are consistent, and the analysis deviation caused by grayscale differences is prevented, ensuring the accuracy of model training and testing; then, the grayscale standardized brain functional magnetic resonance image is transformed into the image with the affine transformation method. The vibration image data is spatially registered with a preset standard template. This is because the brains of different patients may differ in space (such as head position, brain size, etc.). Registration can align the brains of different individuals into a unified spatial coordinate system to ensure accurate matching of anatomical structures, providing a consistent spatial framework for the atrophy assessment of small structures such as the hippocampus, avoiding assessment errors caused by inconsistent spatial positions. Finally, the deviation of low-frequency signals caused by uneven magnetic fields or scanning equipment is removed through deviation field correction. This deviation usually leads to uneven image brightness, affecting image contrast and quality, and obtaining a more uniform image, especially the boundaries between gray matter, white matter and cerebrospinal fluid are clearer, thereby improving image contrast and accuracy, and further enhancing the reliability of subsequent processing steps (such as feature extraction and modeling), thereby improving the accuracy of atrophy assessment.

[0122] Specifically, in step S24, the registered image is subjected to deviation field correction to obtain preprocessed brain functional magnetic resonance image data, which includes the following operation steps:

[0123] Step S241: performing Fourier transform processing on the registered image to obtain a low-frequency image;

[0124] It should be noted that the above-mentioned embodiment of the present application processes the registered image through Fourier transform, separates the low-frequency components from the high-frequency components of the registered image, and then filters out the noise interference of the high-frequency components, retains the low-frequency components (i.e., the above-mentioned low-frequency image), and simplifies the subsequent deviation field estimation operation.

[0125] Step S242: performing an iterative optimization processing operation using the N4ITK algorithm based on the low-frequency image as an initial estimate to obtain preprocessed brain functional magnetic resonance image data.

[0126] It should be noted that the above-mentioned embodiment of the present application first captures global low-frequency changes through Fourier transform processing to provide a preliminary bias field estimate, and then captures local details and inhomogeneities through the iterative optimization processing operation of the N4ITK algorithm. The Fourier transform is combined with the iterative processing of the N4ITK algorithm to reduce the number of iterations of the traditional N4ITK algorithm processing and improve the convergence speed of the algorithm; the above-mentioned N4ITK algorithm (N4ITKBiasFieldCorrection) is an algorithm for medical image processing, in particular, for removing the influence of the "bias field" (BiasField) in the image. The bias field is a low-frequency intensity inhomogeneity that usually appears in medical images, especially in MRI (magnetic resonance imaging) images. This inhomogeneity may be caused by various factors, such as hardware limitations of the scanner, magnetic field inhomogeneities, etc.

[0127] The above-mentioned functional MRI image data of the brain usually produce a low-frequency, gradual intensity inhomogeneity due to reasons such as magnetic field inhomogeneity and detector response differences, which is called bias field. The bias field usually manifests itself as uneven brightness of the image, which will affect the quantitative analysis of the image and subsequent processing tasks. Therefore, bias field correction is an important step to improve image quality and subsequent analysis accuracy. The main goal of bias field correction is to remove or reduce this low-frequency, uneven brightness change from the image, so that the image can more accurately reflect the true physical characteristics of the tissue. In particular, in subsequent image analysis tasks (such as tissue segmentation, quantitative analysis, etc.), removing the bias field can significantly improve the accuracy of the analysis results.

[0128] Specifically, Figure 5 As shown, in step S242, based on the low-frequency image as an initial estimate, an iterative optimization processing operation is performed through the N4ITK algorithm to obtain pre-processed brain functional magnetic resonance image data, including the following operation steps:

[0129] Step S2421: Initialize iteration parameters;

[0130] The iteration parameters include a low-frequency image L(i,j), a deviation field matrix B(i,j), a B-spline grid spacing d, a regularization weight coefficient λ, a spline basis function matrix W(i,j), a learning rate α, and an energy numerical function E(i,j);

[0131] The energy numerical function E(i,j) is calculated as follows:

[0132]

[0133] Where N and M are the number of spline basis functions and bias fields used for calculation, respectively; L(i,j) is the low-frequency image; B(i,j) is the bias field matrix; d is the B-spline grid spacing; λ is the regularization weight coefficient; W(i,j) is the spline basis function matrix;

[0134] It should be noted that the above-mentioned embodiment of the present application obtains the coordinates of each pixel point of the low-frequency image as the "low-frequency image" in the initial iteration parameter, which is expressed in a matrix form, and each element in the matrix represents the coordinates of the pixel point of the low-frequency image; the deviation field matrix represents the deviation field of the initial low-frequency image, and when initially set, it represents the unit matrix; the corrected image is also expressed in a matrix form, which is initially equal to the low-frequency image;

[0135] Step S2422: Calculate based on the low-frequency image L(i,j) and the deviation field matrix B(i,j) to obtain a first corrected image C(i,j);

[0136] It should be noted that the above embodiment of the present application corrects the low-frequency image by applying the initial deviation field to obtain a first version of the corrected image, and adjusts the brightness and contrast of the image through the initial deviation field correction matrix to make it more consistent. The generated first corrected image to be determined provides a preliminary reference for local optimization in subsequent steps;

[0137] Step S2423: searching the first corrected image to be determined based on a search window of a preset size to obtain a local area image matrix; calculating a local energy value E based on the local area image matrix and the energy value function local(i,j) ;

[0138] It should be noted that the above-mentioned embodiments of the present application identify local deviations and inconsistencies in the image through local calculations of local areas and feedback of energy numerical functions, provide local energy numerical feedback for iterative optimization, and guide the optimization process to adjust the image in a more fine-grained manner.

[0139] Step S2424: Determine the local energy value E local(i,j)Is it greater than a preset local energy value minimum threshold value P; if so (if not, it is determined that the local area image matrix is ​​an optimized local area image matrix), then the local energy value E is optimized based on the learning rate α. local(i,j) The corresponding local area image matrix is ​​locally iteratively optimized to obtain an optimized local area image matrix;

[0140] It should be noted that the above-mentioned embodiment of the present application uses the learning rate to adjust the optimization step size, and refines the local area through local iteration to achieve higher image quality. Through local optimization, the image gradually approaches the real correction effect in the detail area.

[0141] Step S2425: constructing a second corrected image to be determined based on the optimized local image matrix;

[0142] Step S2426: performing bias field estimation based on the second corrected image to be determined and the low-frequency image L(i,j) to obtain an updated bias field matrix B'(i,j);

[0143] It should be noted that the above-mentioned embodiment of the present application can gradually eliminate uneven brightness and artifacts in the image by continuously adjusting the deviation field, and the updated deviation field provides a more accurate correction basis for the next iteration.

[0144] Step S2427: Based on the second corrected image to be determined and the updated deviation field matrix B'(i,j), a corrected energy value E'(i,j) is calculated by using an energy value function E(i,j);

[0145] It should be noted that, in the above-mentioned embodiment of the present application, the second corrected image to be determined is used as the low-frequency image L(i, j), and the updated deviation field matrix B'(i, j) is used as the initial deviation field matrix B(i, j), and the corrected energy value E'(i, j) is calculated by the energy value function E(i, j);

[0146] Step S2428: Determine whether the corrected energy value E'(i,j) is greater than or equal to the preset minimum threshold of the total energy function; if so, output the second corrected image to be determined as the preprocessed brain functional MRI image data; if not, return to the above step S2422 and iterate again until the preprocessed brain functional MRI image data is output.

[0147] It should be noted that the above-mentioned embodiment of the present application gradually removes the uneven brightness and artifact problems in the original image (i.e., the above-mentioned low-frequency image) through updating and local optimization of the deviation field, improves the image quality, and further makes the intensity distribution of the image in the overall and local areas more consistent through regularization terms and local iterative optimization; through the feedback of the energy numerical function in the iterative process, the gradual optimization of the image is ensured, and finally the preset correction effect is achieved to avoid overfitting;

[0148] The above-mentioned embodiment of the present application first provides the necessary initial conditions for the subsequent deviation field correction by initializing the parameters, ensuring that the iterative process has a clear starting point and can continuously approach the goal (removing the deviation field), and the initialization of the low-frequency image and the deviation field matrix provides a basis for subsequent optimization and iteration, so that the subsequent steps can eliminate the unevenness in the image by gradual correction; then, by preliminarily applying the deviation field matrix to correct the low-frequency image, a first version of the corrected image is obtained, the low-frequency uneven brightness changes are removed, and a preliminary unified brightness correction effect is provided, which provides a basis for subsequent local optimization, especially when the details and local areas in the functional magnetic resonance imaging of the brain have more complex deviations, more refined adjustments can be made through subsequent iterations; further, by selecting a local area in the image and calculating the local energy value based on the energy numerical function, the deviation and inconsistency in the local area are identified, and through the optimization of the local area, the deviation field correction is not limited to global adjustment, but can also finely process the local brightness unevenness, which helps to identify subtle and irregular deviations in the image, thereby improving the correction accuracy of the image, especially for small-range areas in the functional magnetic resonance imaging of the brain. The problem of local brightness unevenness is solved; further, through local iterative optimization, the local area in the image is finely adjusted to gradually eliminate the deviation and brightness unevenness in the functional magnetic resonance imaging of the brain, and the learning rate controls the step size of each iteration to avoid over-adjustment or slow convergence; further, the second version of the corrected image has been locally iteratively optimized, so that the image quality has been further improved in details, and then the image is used as input to update the deviation field matrix and provide an update basis for the next round of correction; further, by comparing the second corrected image to be determined with the low-frequency image, the deviation field matrix is ​​estimated and updated, thereby further improving the brightness uniformity of the image; further, the corrected energy value is calculated by the energy numerical function to evaluate the correction effect of the deviation field and provide a basis for the next step to determine whether the correction is completed; by judging whether the deviation field correction reaches the predetermined accuracy threshold, if it reaches it, the final preprocessed functional magnetic resonance imaging of the brain is output, otherwise, the iteration step is returned to continue the optimization until the deviation field problem of the image is completely solved, and the corrected image finally output can accurately reflect the real physical characteristics of the brain and avoid the influence of the brightness unevenness caused by the deviation field on the subsequent analysis.

[0149] Specifically, in step S2424, the local energy value E is calculated based on the learning rate α. local(i,j) The corresponding local area image matrix is ​​locally iteratively optimized to obtain an optimized local area image matrix, including the following steps:

[0150] Step S24241: Using the gradient descent method to calculate the local area image matrix Q based on the learning rate local(i,j) Perform local optimization to obtain a first optimized image matrix;

[0151] The first optimized image matrix is ​​calculated as follows:

[0152]

[0153] In the formula, Represents the local energy value E local(i,j) The partial derivative of the pixel value of each pixel point of the current local area image matrix (indicating the influence of the pixel point on the first optimized image matrix);

[0154] Step S24242: Calculate the energy of the first optimized image matrix based on the energy value function E(i,j) to obtain a first local energy value E' local(i,j) ;

[0155] Step S24243: Determine the first local energy value E' local(i,j) Is it greater than the preset local energy value minimum threshold value P? If not, then output the first optimized image matrix as the optimized local area image matrix; if so, then further determine the first local energy value E' local(i,j) Whether the energy ratio to the local energy value minimum threshold value P is greater than or equal to the preset step size adjustment threshold; if so (if not, directly return to the above step S24241 and iterate again), then adjust the learning rate α based on the energy ratio to obtain the adjusted first learning rate α';

[0156] The first learning rate α' is calculated as follows:

[0157]

[0158] Among them, β is the step size adjustment coefficient (controls the sensitivity of learning rate adjustment);

[0159] Step S24244: returning the first learning rate α' and the first optimized image matrix to the above step S24241 for re-iteration until an optimized local area image matrix is ​​output.

[0160] It should be noted that the above-mentioned embodiment of the present application first optimizes the local area of ​​the image by the gradient descent method to reduce the energy function value of the image, optimize the deviation and brightness unevenness; further, the "quality" of the optimized image in the current local area is quantified by calculating the energy value of the updated image matrix; further, by comparing the first local energy value with the preset minimum threshold, it is determined whether the current optimization has met the requirements. If the requirements are not met, it means that the optimization degree of the image is insufficient, and there is still a large brightness unevenness or deviation field problem, and it is necessary to continue iterative optimization; further, if the first local energy value is greater than Minimum threshold, then compare whether the ratio of the first local energy value to the minimum threshold is greater than the preset step adjustment threshold. If the ratio is large, it means that the optimization step is too large, which may lead to unstable optimization or over-adjustment. Therefore, the learning rate needs to be adjusted to obtain the adjusted first learning rate (α'); when it is determined that the optimization needs to continue (that is, the energy value does not reach the minimum threshold), the learning rate is adjusted according to the energy ratio. The adjusted learning rate will affect the step size of the next iteration to avoid the step size being too large or too small. The adjusted learning rate enables each iteration to better control the degree of optimization of image details during local optimization.

[0161] The method for identifying and processing abnormal atrophy of the hippocampus in functional magnetic resonance imaging of the brain provided by the present invention innovatively introduces an age stratification standard and establishes a normal value range database based on age. The MTA score calculation adopts an improved automatic scoring algorithm to convert qualitative evaluation into quantitative indicators. Studies have found that the above-mentioned dynamic evaluation method in the embodiment of the present invention can better monitor the progression of the disease, and at the same time, by establishing a standardized scoring system, the comparability of results between different centers is improved. The evaluation model provided by the present invention has been verified by clinical trials and the results show that the overall diagnostic judgment accuracy rate reaches 89.5% to 90% (also the model judgment accuracy rate). The above-mentioned model judgment accuracy rate indicates that the model has high reliability in clinical applications.

[0162] In summary, the present invention proposes a method for identifying and processing abnormal atrophy of the hippocampus in a functional magnetic resonance imaging (MRI) image of the brain, which collects the patient's functional magnetic resonance imaging (FMRI) image data containing the functional activities, structure and potential pathological information of the brain through medical imaging equipment (such as magnetic resonance imaging equipment), so as to provide original input for subsequent analysis; further, by performing preprocessing operations such as bias field correction processing on the collected image data, high-quality input is provided for subsequent feature extraction and modeling; then, the hippocampus region is extracted from the preprocessed image data, and a three-dimensional image of the hippocampus is created based on the region, and the hippocampus region is accurately located and extracted, and then three-dimensional reconstruction is performed using this information to present the morphological characteristics of the hippocampus in more detail, especially when evaluating changes in the volume of the hippocampus (such as atrophy), which is more advantageous; further, a deep learning network (such as a multi-scale convolutional neural network) is used to perform feature extraction on the three-dimensional hippocampus image. The shape, volume and texture features of the hippocampus are obtained, and the adaptability of the model to changes under different image conditions is improved by combining a multi-scale convolutional neural network with an attention mechanism, thereby improving the accuracy and comprehensiveness of feature extraction and more accurately reflecting the pathological characteristics of the hippocampus. Finally, based on the extracted features, the abnormal atrophy of the hippocampus is classified and identified, and a classification result (such as normal or abnormal) is generated, which provides an effective decision support tool for the clinic and assists doctors in formulating treatment plans. Moreover, on the basis of the above scheme, age information is introduced to provide a physiological background for the assessment of hippocampal atrophy, because normal hippocampal atrophy is closely related to age, especially in the elderly. Combined with age information, it can be compared with normal reference values ​​based on age stratification to determine whether the subject's hippocampus has abnormal atrophy, thereby avoiding misdiagnosis or misevaluation caused by ignoring age differences in traditional hippocampal atrophy analysis and processing methods.

[0163] Specifically, in the preprocessing process of the deviation field correction, the global low-frequency changes are captured through Fourier transform processing to provide a preliminary deviation field estimate. Then, the local details and inhomogeneities are captured through the iterative optimization processing of the N4ITK algorithm. The combination of Fourier transform and N4ITK algorithm iterative processing reduces the number of iterations of the traditional N4ITK algorithm processing and improves the convergence speed of the algorithm, thereby outputting the preprocessed brain functional magnetic resonance images more quickly and accurately, and improving the subsequent hippocampal atrophy identification analysis.

[0164] Furthermore, through local area optimization, the deviation field correction is not limited to global adjustment, but can also finely process local brightness unevenness, which helps to identify subtle and irregular deviations in the image, thereby improving the correction accuracy of the image;

[0165] In addition, the local area of ​​the image is optimized by the gradient descent method, and the learning rate is adjusted to affect the step size of the next iteration to avoid too large or too small a step size. The adjusted learning rate enables each iteration to better control the degree of optimization of image details during local optimization, thereby obtaining more accurate preprocessed brain functional MRI images and improving subsequent hippocampal atrophy identification and analysis.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. A person skilled in the art may modify the technical solutions described in the above embodiments, or replace part or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying and processing abnormal atrophy of the hippocampus in brain functional magnetic resonance imaging, characterized in that: The steps are as follows: Acquiring brain functional magnetic resonance imaging data to be processed; Preprocessing the brain functional magnetic resonance image data to obtain preprocessed brain functional magnetic resonance image data; Extracting the hippocampus region from the preprocessed brain functional magnetic resonance image data; Establishing a three-dimensional image of the hippocampus based on the hippocampus region; Performing feature extraction on the three-dimensional image of the hippocampus region based on a deep learning network to obtain a feature extraction result; According to the feature extraction result, the abnormal atrophy classification and identification of the hippocampus region is performed to obtain a classification and identification result.

2. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 1, characterized in that: After obtaining the brain functional magnetic resonance imaging data to be processed, the following steps are also included: Obtaining age information of a subject corresponding to the brain functional magnetic resonance imaging data to be processed; Extracting hippocampus region data in the coronal T1-weighted image in the hippocampus three-dimensional image; Establishing a quantitative assessment model for the degree of hippocampal atrophy based on the hippocampal region data; The age information A, the quantitative evaluation model H for the degree of hippocampal atrophy and the feature extraction result F are combined to calculate the grading of hippocampal atrophy and obtain a grading result S.

3. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 2, characterized in that: The hippocampus region data includes hippocampus volume data and morphological feature parameter data; The quantitative assessment model for the degree of hippocampal atrophy includes a hippocampal atrophy rate calculation model, a hippocampal volume loss speed calculation model, and a normal reference value database based on age stratification.

4. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 3, characterized in that: The step of combining the age information A, the quantitative evaluation model H for the degree of hippocampal atrophy and the feature extraction result F to calculate the hippocampal atrophy grading determination to obtain the grading determination result S includes the following steps: Based on the age information A, the normal reference value data corresponding to the age information A is retrieved from the normal reference value database based on age stratification; A quantitative evaluation model H' for the degree of hippocampal atrophy after retrieval is established based on the normal reference value data corresponding to the age information A, the hippocampal atrophy rate calculation model, and the hippocampal volume loss speed calculation model; The age information A, the quantitative evaluation model H' for the degree of hippocampal atrophy after retrieval, and the feature extraction result F are standardized to obtain standardized age information A', the standardized quantitative evaluation model H' for the degree of hippocampal atrophy after retrieval, and the standardized feature extraction result F'; Calculating the graded judgment result S based on the standardized age information A', the standardized post-retrieval hippocampal atrophy degree quantitative assessment model H", and the standardized feature extraction result F', to obtain the graded judgment result S; The calculation method of the classification judgment result S is: S=w1×A′+w2×H”+w3×F′+b; Wherein, w1, w2 and w3 are weight coefficients respectively; b is the bias term; A' is the standardized age information; H" is the standardized quantitative assessment model of hippocampal atrophy; and F' is the standardized feature extraction result.

5. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 4, characterized in that: The preprocessing of the brain functional nuclear magnetic resonance image data to obtain the preprocessed brain functional nuclear magnetic resonance image data comprises the following steps: The non-local mean filtering algorithm is used to denoise the brain functional magnetic resonance image data to obtain denoised brain functional magnetic resonance image data; Performing grayscale value standardization on the denoised brain functional magnetic resonance image data to obtain grayscale standardized brain functional magnetic resonance image data; The grayscale-normalized functional magnetic resonance imaging data of the brain are spatially registered with a preset standard template using an affine transformation method to obtain a registered image. The registered images are corrected for deviation fields to obtain preprocessed brain functional magnetic resonance image data.

6. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 5, characterized in that: The method of performing deviation field correction on the registered image to obtain pre-processed brain functional magnetic resonance image data includes the following steps: Performing Fourier transform processing on the registered image to obtain a low-frequency image; Based on the low-frequency image as an initial estimate, an iterative optimization processing operation is performed through the N4ITK algorithm to obtain pre-processed brain functional magnetic resonance imaging data.

7. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 6, characterized in that: The method of performing iterative optimization processing based on the low-frequency image as an initial estimate through the N4ITK algorithm to obtain pre-processed brain functional magnetic resonance image data includes the following steps: Initialize iteration parameters; The iteration parameters include a low-frequency image L(i,j), a deviation field matrix B(i,j), a B-spline grid spacing d, a regularization weight coefficient λ, a spline basis function matrix W(i,j), a learning rate α, and an energy numerical function E(i,j); Calculating based on the low-frequency image L(i,j) and the deviation field matrix B(i,j) to obtain a first corrected image C(i,j); Based on a search window of a preset size, the first corrected image to be determined is searched to obtain a local area image matrix; based on the local area image matrix and the energy value function, a local energy value E is calculated. local(i,j) ; Determine the local energy value E local(i,j) Is it greater than the preset local energy value minimum threshold P? If so, the local energy value E is adjusted based on the learning rate α. local(i,j) The corresponding local area image matrix is ​​locally iteratively optimized to obtain an optimized local area image matrix; Constructing a second corrected image to be determined based on the optimized local image matrix; Performing deviation field estimation based on the second corrected image to be determined and the low-frequency image L(i,j) to obtain an updated deviation field matrix B'(i,j); Based on the second corrected image to be determined and the updated deviation field matrix B'(i,j), a corrected energy value E'(i,j) is calculated by using an energy value function E(i,j); Determine whether the corrected energy value E'(i,j) is greater than or equal to the preset minimum threshold of the total energy function; if so, output the second corrected image to be determined as the preprocessed brain functional magnetic resonance image data; if not, return to the above step S2422 and iterate again until the preprocessed brain functional magnetic resonance image data is output.

8. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 7, characterized in that: The energy numerical function E(i,j) is calculated as follows: Where N and M are the number of spline basis functions and bias fields used for calculation, respectively; L(i,j) is the low-frequency image; B(i,j) is the bias field matrix; d is the B-spline grid spacing; λ is the regularization weight coefficient; W(i,j) is the spline basis function matrix.

9. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 8, characterized in that: The local energy value E is calculated based on the learning rate α. local(i,j) The corresponding local area image matrix is ​​locally iteratively optimized to obtain an optimized local area image matrix, including the following steps: Based on the learning rate, the local area image matrix Q is calculated by using the gradient descent method. local(i,j) Perform local optimization to obtain a first optimized image matrix; The energy of the first optimized image matrix is ​​calculated based on the energy value function E(i, j) to obtain a first local energy value E' local(i,j) ; Determine the first local energy value E' local(i,j) Is it greater than the preset local energy value minimum threshold value P? If not, the first optimized image matrix is ​​output as the optimized local area image matrix; if so, the first local energy value E' is further determined. local(i,j) whether the energy ratio to the local energy value minimum threshold value P is greater than or equal to the preset step size adjustment threshold; if so, adjusting the learning rate α based on the energy ratio to obtain an adjusted first learning rate α'; The first learning rate α' and the first optimized image matrix are returned to the above operation for re-iteration until an optimized local area image matrix is ​​output.

10. The method for identifying and processing abnormal atrophy of the hippocampus in a cranial functional magnetic resonance imaging according to claim 9, characterized in that: The first optimized image matrix is ​​calculated as follows: In the formula, Represents the local energy value E local(i,j) Partial derivatives of the pixel values ​​of each pixel point of the image matrix in the current local area; The first learning rate α' is calculated as follows: Among them, β is the step size adjustment coefficient.

Images