Multi-Resolution Graph Analysis of Dynamic Brain Network for Classification of Alzheimer's Disease and Mild Cognitive Impairment

2024-09-06

AI Summary

Early diagnosis of Alzheimer’s disease (AD) and mild cognitive impairment (MCI) remains challenging due to clinical and neurobiological heterogeneity. To address this, we propose a novel multi-scale dynamic brain network framework integrating discrete wavelet transform (DWT) with graph-theoretic analysis, overcoming limitations of static functional connectivity modeling. Our approach constructs dynamic functional connectivity graphs from resting-state fMRI data, extracts multi-scale topological graph features across frequency bands, and employs SVM and XGBoost for three-way classification (AD, MCI, healthy controls). For the first time, it identifies AD/MCI-specific time-frequency functional connectivity abnormalities across multiple frequency bands. Evaluated on the ADNI dataset, the method achieves 89.2% classification accuracy and precisely localizes aberrant connections within critical networks—including the default mode network—in disease-relevant frequency bands. This interpretable, generalizable computational neuroimaging framework advances early, differential, and longitudinal assessment of AD progression.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Humans and AI: Brain-Sensing and Analysis
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Machine learning and data science for the Web
• Social Networks and Social Media: Computational social science

Abstract

Alzheimer's disease (AD) is a neurodegenerative disorder marked by memory loss and cognitive decline, making early detection vital for timely intervention. However, early diagnosis is challenging due to the heterogeneous presentation of symptoms. Resting-state functional magnetic resonance imaging (rs-fMRI) captures spontaneous brain activity and functional connectivity, which are known to be disrupted in AD and mild cognitive impairment (MCI). Traditional methods, such as Pearson's correlation, have been used to calculate association matrices, but these approaches often overlook the dynamic and non-stationary nature of brain activity. In this study, we introduce a novel method that integrates discrete wavelet transform (DWT) and graph theory to model the dynamic behavior of brain networks. Our approach captures the time-frequency representation of brain activity, allowing for a more nuanced analysis of the underlying network dynamics. Machine learning was employed to automate the discrimination of different stages of AD based on learned patterns from brain network at different frequency bands. We applied our method to a dataset of rs-fMRI images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, demonstrating its potential as an early diagnostic tool for AD and for monitoring disease progression. Our statistical analysis identifies specific brain regions and connections that are affected in AD and MCI, at different frequency bands, offering deeper insights into the disease's impact on brain function.

Problem

• Classifying Alzheimer's disease stages using multi-scale fMRI analysis
• Capturing dynamic brain network dynamics overlooked by traditional methods
• Identifying frequency-specific brain region alterations in AD/MCI

Innovation

• Integrated discrete wavelet transform with graph theory
• Modeled dynamic brain network behavior across frequencies
• Used machine learning for automated AD stage classification