基于机器学习和脑网络的阿尔茨海默病多模态影像研究进展

阿尔茨海默病是发生在中老年人群体中的神经退行性疾病，以记忆障碍和认知能力下降为主要特征。目前大多数关于阿尔茨海默病的多模态影像研究主要集中在局部特定脑区，缺乏对全脑网络模式的深入探讨。本文针对基于机器学习和脑网络的阿尔茨海默病多模态影像研究进行概述。首先，介绍了阿尔茨海默病的定义以及机器学习技术在脑疾病影像研究中的局限性；其次，阐述了机器学习在脑网络预测中的通用流程，主要包括：特征提取、特征选择与特征降维、模型构建、模型评价；最后，依次介绍了机器学习在阿尔茨海默病的灰质结构脑网络、白质结构脑网络、静息态功能脑网络以及多模态融合脑网络的研究成果。通过对近年来研究成果的梳理，本文对该领域未来发展方向进行了以下三点展望：大样本多中心研究，具有可解释性的深度学习技术，建立纵向预测模型。

Advances in multimodal imaging of Alzheimer's disease based on machine learning and brain network

Alzheimer's disease is a neurodegenerative disease occurring in middle-aged and elderly people, characterized by memory impairment and cognitive decline. At present, most of the multimodal imaging studies on Alzheimer's disease mainly focus on local specific brain regions, and there is a lack of in-depth exploration of whole brain network patterns. This paper summarizes the research on multimodal imaging of Alzheimer's disease based on machine learning and brain network. First, it introduces the definition of Alzheimer's disease and the limitations of machine learning technology in brain disease imaging research. Secondly, the general process of machine learning in brain network prediction is described, including: feature extraction, feature selection and feature dimensionality reduction, model construction and model evaluation; Finally, the research results of machine learning in Alzheimer's disease brain network of gray matter structure, white matter structure, resting-state functional brain network and multimodal fusion brain network are introduced in turn. By combing the research results in recent years, this paper has made the following three prospects for the future development direction of this field: large sample multi- center research, interpretable deep learning technology, and the establishment of longitudinal prediction models.