Alzheimer's Disease Diagnostic Method Based on Adaptive Fine-Grained Causal Brain Networks

Abstract

This invention discloses an Alzheimer's disease diagnostic method based on an adaptive fine-grained causal brain network, belonging to the field of bioinformatics. It mainly comprises four parts: adaptive fine-grained segmentation, causal brain network construction, intervention-based causal network fusion, and disease diagnosis based on the causal brain network. The specific steps are as follows: First, an adaptive fine-grained segmentation method is used to finely segment the brain network according to the temporal periodic differences of brain regions, capturing spatiotemporal features at different scales. Then, spatiotemporal features are extracted from these fine-grained segments, and feature learning is performed using deep learning and frequency domain algorithms to obtain multi-granular feature representations of the brain network. Subsequently, causal inference techniques are used to construct a causal brain network, identifying causal relationships between brain regions and analyzing the correlations between different brain regions. Furthermore, causal intervention strategies are combined to integrate information from the brain region-level causal network into the fine-grained causal network. Finally, by constructing a diagnostic model based on the fine-grained causal brain network and using graph neural networks to classify brain network features, early diagnosis of Alzheimer's disease is achieved. This process allows for the construction of causal brain networks based on brain region segmentation, revealing causal changes between brain regions and improving the diagnostic results of Alzheimer's disease.

Description

Technical Field

[0001] This invention relates to the field of bioinformatics, and more specifically to a diagnostic method for Alzheimer's disease based on an adaptive fine-grained causal brain network. Background Technology

[0002] Alzheimer's disease (AD) is one of the most prevalent neurodegenerative diseases, and recent advances in deep learning have demonstrated the immense potential of neuroimaging data analysis in diagnosing this type of disease. Traditional methods typically rely on static and linear models, such as those based on convolutional neural networks or Transformers, but these may fail to effectively capture the complex, dynamic, and nonlinear properties of brain connectivity. Therefore, researchers have naturally considered introducing graph structures to model brain networks. In graph neural network (GNN)-based approaches, Shimizu applied LiNGAM, a method focused on estimating non-Gaussian causal structures, providing a framework for revealing causal relationships in brain networks. However, while complex, these methods often neglect cyclic temporal dynamics and causal dependencies, which are crucial for a comprehensive understanding of the neurodegenerative processes behind attention deficit disorder. Li et al. proposed BrainGNN—an interpretable brain graph neural network specifically designed for fMRI analysis—to address some of these challenges. BrainGNN plays a significant role in identifying linear causal structures in brain networks, but it still faces difficulties in handling the complexity of nonlinear interactions and the high-dimensional data typical of fMRI studies. Therefore, there is still an urgent need for a comprehensive approach that combines deep learning with causal reasoning to more accurately model the dynamic nonlinear interactions in neurodegenerative diseases such as Alzheimer's disease (AD). Summary of the Invention

[0003] In view of this, the present invention provides an Alzheimer's disease diagnostic method based on an adaptive fine-grained causal brain network to solve the problem of combining temporal dynamics and causal dependencies for the analysis and diagnosis of Alzheimer's disease. This method is helpful for the research of brain network construction methods, thereby better discovering and diagnosing Alzheimer's disease. This is related to the early diagnosis and intervention of Alzheimer's disease and is of great significance for slowing disease progression and alleviating the worsening of symptoms.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: an Alzheimer's disease diagnostic method based on an adaptive fine-grained causal brain network, the specific steps of which include the following:

[0005] A flowchart illustrating the process of diagnosing Alzheimer's disease using an adaptive fine-grained causal brain network-based approach.

[0006] Data Collection and Preprocessing: First, real sample datasets of functional magnetic resonance imaging (fMRI) for Alzheimer's disease were collected; data preprocessing included denoising, removing irrelevant signals, etc., and the image data was processed into time series data for subsequent analysis and modeling;

[0007] Adaptive fine-grained segmentation: In brain network analysis, a time-series periodic decoupling method is used to adaptively segment different brain regions into fine-grained time series.

[0008] Feature extraction at different granularities: Based on adaptive fine-grained segmentation, information at different granularities is extracted. This includes extracting data features of neural activity at multiple levels, including brain region level and time slice level: Deep learning and frequency domain methods are used to perform multi-level feature learning on high-dimensional data to capture the dynamic changes and structural features of brain networks;

[0009] Causal brain network construction: Using the extracted multi-granularity data features, multi-granularity causal variables are established. The Granger causality discovery method is used to identify causal relationships between different brain regions and analyze the causal direction and degree of influence between brain regions or time slices.

[0010] Intervention-based causal network fusion: Based on the construction of causal brain networks of different granularities, and combined with the basic principles of causal graph intervention, the information of coarse-grained causal brain networks at the brain region level is integrated into the fine-grained causal brain network at the time slice level to complete the final fine-grained causal brain network construction.

[0011] Disease diagnosis based on causal brain networks: Alzheimer's disease is diagnosed based on graph neural networks by utilizing a constructed fine-grained causal brain network and multi-level features.

[0012] 2. Furthermore, the main periodicity of the time series is generated using periodic pattern decoupling and the TopM method, for a single brain region of a sample. In general, the value of the main cycle is: , .

[0013] 3. Further, in the selection of segmentation points for the fine-grained segmentation, for each brain region, the principal period of the time series and its multiples are the segmentation points, and the value of the principal period is: .

[0014] 4. Furthermore, the fusion of the causal networks utilizes a causal graph intervention method, intervening in each causal variable of the coarse-grained brain network by removing all its incoming edges, and then calculating the difference in causal effects before and after the intervention: Then the intervention results This is added to fine-grained causal brain networks to enhance them.

[0015] As can be seen from the above technical solution, compared with the prior art, the present invention discloses an Alzheimer's disease diagnosis method based on an adaptive fine-grained causal brain network. The beneficial effects of the present invention are as follows:

[0016] 1. This invention proposes an Alzheimer's disease diagnosis method based on an adaptive fine-grained causal brain network, which can comprehensively improve the performance of fMRI diagnosis of Alzheimer's disease by utilizing graph deep learning and the causal discovery capabilities at the brain region level and time segment level.

[0017] 2. Based on the periodic decoupling method of time series, this invention performs adaptive partitioning and embedded representation of fine-grained causal variables, which is conducive to a more thorough time-segment level understanding of the causal dynamics of neurological diseases in the time dimension. Compared with existing Alzheimer's disease diagnosis methods, it has a simpler and more interpretable embedded representation.

[0018] 3. The causal graph fusion model proposed in this invention integrates coarse-grained causal graphs into fine-grained brain networks through intervention operations on the causal graphs, thereby effectively enhancing the representation of macroscopic features in the brain network structure in an interpretable manner. Attached Figure Description

[0019] 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.

[0020] Figure 1 A diagram illustrating the diagnostic process for Alzheimer's disease based on an adaptive fine-grained causal brain network;

[0021] Figure 2 This is the overall flowchart of the present invention; Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] This invention discloses a diagnostic method for Alzheimer's disease based on an adaptive fine-grained causal brain network, the specific steps of which include the following:

[0024] Step 1: Data Collection and Preprocessing. Collect real fMRI data on Alzheimer's disease. First, preprocess the raw fMRI datasets of N subjects into time series data. , representing a coarse-grained time series at the brain region level. Each subject is characterized by K brain regions spanning time T, denoted as . ,in Let t be the time point in the k-th brain region of the i-th subject. Unlike coarse-grained representation, fine-grained representation of time series involves segmenting time segments within the same brain region, which contain different lengths of time depending on the period. In the brain's effective functional connectivity network, these fine-grained time segments are treated as causal variables, represented as... ,in This represents the a-th time segment obtained from the time series that corresponds to the feature k of the i-th subject.

[0025] Step 2: Adaptive fine-grained segmentation. For the time series of each brain region... All samples are projected into the Ramanujan subspace, thus achieving periodic pattern decomposition. Then, the top M most frequent main cycles in each brain region sample are selected as the segmentation points for that cycle. This process allows for adaptive segmentation of fine-grained time segments without manual parameter setting. The value of M is determined based on the observed dataset S, as follows: , .in, This indicates the number of major periodic components in the time series extracted from brain region k by the APMD method for sample i. This represents the average number of periodic components across all samples within brain region k. Once the principal periodic components of sequence S are determined, each extracted period is systematically stored in a pre-designed three-dimensional tensor. In the first dimension, represents the m-th major cycle extracted from the time series of feature k in sample i. The second dimension, , maps to the various features considered, capturing the cyclical behavior of each specific brain region in the dataset. The third dimension, , encompasses the different cycles determined for each feature to reflect the multi-scale cyclical dynamics in the time series. Specifically, for each brain region, the cycle and its multiples are determined as the segmentation points of the time series in that specific brain region: , Among them, C k It is the main periodic set of brain region k.

[0026] Step 3: Feature Extraction at Different Granularities. Wavelet transforms were applied to fine-grained time segments and coarse-grained time series for each brain region. By projecting the data from the time domain to the frequency domain, this step allows us to extract complex temporal patterns into better feature components. Frequency domain representation enhances our ability to detect subtle oscillatory dynamics and temporal variations that might be masked in the time domain. The result of this process is two types of embeddings that capture both causal variables and preserve the original order of the time series. One is the fine-grained time segment embedding, which dissects local adaptive periodic features; the other is the coarse-grained embedding, which dissects global macroscopic features of the brain region.

[0027] Step 4: Causal Brain Network Construction. To independently learn coarse-grained and fine-grained causal brain networks, this stage employs a graph encoder to derive a dynamic adjacency matrix. This network utilizes the Granger causal discovery algorithm to analyze time series and discover fine-grained and coarse-grained causal graphs to reflect the potential fine-grained causal relationships between brain structures.

[0028] Step 5: Intervention-Based Causal Network Fusion. Graph Isomorphic Networks (GINs) are used to enhance the edge weights of the fine-grained causal brain network by leveraging causal information from the coarse-grained network. Specifically, we intervene for each causal variable in the coarse-grained graph, removing all incoming edges, and use the intervened causal graph to calculate the average causal effect (ACE): Then, the ACE will enhance the marginal weights of the fine-grained brain network by incorporating intervention enhancement values: This integrated approach, combining detailed local feature learning with global causal context, enhances the overall understanding of brain networks.

[0029] Step Six: Disease Diagnosis Based on Causal Brain Networks. By inputting the learned causal brain network into the GNN, the model can effectively capture and utilize the nonlinear causal relationships between fine-grained brain regions, thereby making accurate diagnoses.

[0030] The present invention proposes an Alzheimer's disease diagnostic method based on an adaptive fine-grained causal brain network, which solves the problem of combining temporal dynamics and causal dependence for the analysis and diagnosis of Alzheimer's disease. The method is effective and easy to implement.

[0031] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A diagnostic method for Alzheimer's disease based on adaptive fine-grained causal brain networks, characterized in that, The specific steps include the following: Data Collection and Preprocessing: First, real sample datasets of functional magnetic resonance imaging (fMRI) for Alzheimer's disease were collected; data preprocessing included denoising, removing irrelevant signals, and converting the image data into time series data for subsequent analysis and modeling. Adaptive fine-grained segmentation: In brain network analysis, a time-series periodic decoupling method is used to adaptively segment different brain regions into fine-grained time series. Feature extraction at different granularities: Based on adaptive fine-grained partitioning, information at different granularities is extracted, including data features of neural activity at multiple levels, including brain region level and time slice level: deep learning and frequency domain methods are used to learn features at multiple levels on high-dimensional data to capture the dynamic changes and structural features of brain networks; Causal brain network construction: Using the extracted multi-granularity data features, multi-granularity causal variables are established. The Granger causality discovery method is used to identify causal relationships between different brain regions and analyze the causal direction and degree of influence between brain regions or time slices. Intervention-based causal network fusion: Based on the construction of causal brain networks of different granularities, and combined with the basic principles of causal graph intervention, the information of coarse-grained causal brain networks at the brain region level is integrated into the fine-grained causal brain network at the time slice level to complete the final fine-grained causal brain network construction. Disease diagnosis based on causal brain networks: Alzheimer's disease is diagnosed based on graph neural networks by utilizing a constructed fine-grained causal brain network and multi-level features.

2. The Alzheimer's disease diagnostic method based on adaptive fine-grained causal brain networks according to claim 1, characterized in that, The generation method of the main periods of the time series adopts a decoupled periodic pattern combined with the frequency of periodic occurrence, for a brain region of a sample. In general, the value of the main cycle is: , .

3. The Alzheimer's disease diagnostic method based on adaptive fine-grained causal brain networks according to claim 1, characterized in that, The selection of segmentation points for the fine-grained segmentation involves using the principal period and its multiples in the time series as segmentation points for each brain region. The value of the principal period is: .

4. The Alzheimer's disease diagnostic method based on adaptive fine-grained causal brain networks according to claim 1, characterized in that, The fusion of the causal networks utilizes a causal graph intervention method to intervene in each causal variable of the coarse-grained brain network. All incoming edges of the intervened nodes in the coarse-grained network are removed, and then the difference in causal effect before and after the intervention is calculated. Then the intervention results This is added to fine-grained causal brain networks to enhance them.

Images