Year: 2025
Type: article
Abstract: 在人工智能技术飞速发展的当下，病理医学领域数据展现出多模态、高维度及复杂关联的特性，传统机器学习方法处理这类非欧几里得结构数据时困难重重。图神经网络(GNNs)作为新兴深度学习模型，能够高效处理图结构数据并获取深层次特征。医学领域中蛋白质相互作用网络、基因调控网络等应用场景与图结构天然适配，促使GNN模型在医学数据分析中的应用成为人工智能与医学交叉研究的热门方向。本文深入剖析近年来GNN模型在医学领域数据的应用，不仅阐述了其基本原理和模型架构，还从脑网络分析、疾病诊断与预测、药物发现与相互作用预测以及模型可解释性等多个维度，对相关研究进行详细解读与深度分析，提出模型设想，探讨未来发展方向与面临的挑战。目的在于为医学和人工智能领域的研究者提供全面且系统的参考，助力GNN模型在医学领域实现更广泛、更深入的应用与发展。In the current era of the rapid development of artificial intelligence technology, the data in the medical field exhibit characteristics of multimodality, high-dimensionality, and complex correlations. Traditional machine learning methods face numerous difficulties when dealing with this non-Euclidean structured data. As an emerging deep-learning model, Graph Neural Networks (GNNs) can efficiently process graph-structured data and obtain deep-level features. Application scenarios in the medical field, such as protein-protein interaction networks, gene regulatory networks, etc., are naturally compatible with graph structures. This has made the application of GNN models in medical data analysis a hot topic in the interdisciplinary research of artificial intelligence and medicine. This paper deeply analyzes the applications of GNN models in medical field data in recent years. It not only elaborates on their basic principles and model architectures but also, from multiple dimensions including brain network analysis, disease diagnosis and prediction, drug discovery and interaction prediction, and model interpretability, conducts detailed interpretations and in-depth analyses of relevant research, and explores future development directions and challenges. The aim is to provide a comprehensive and systematic reference for researchers in both the medical and artificial intelligence fields, and to facilitate the more extensive and in-depth application and development of GNN models in the medical field.
Author: 敏 鞠
Institution: University of Shanghai for Science and Technology
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4409074397