Page 53 - 《中国医疗器械杂志》2026年第2期
P. 53
Chinese Journal of Medical Instrumentation 2026年 第50卷 第2期
综 合 评 述
文章编号:1671-7104(2026)02-0167-14
类脑人工神经网络的多维度建模及其
在医学影像分析中的应用
1,2
【作 者】 王采薇 ,陈思霖 ,蒋希 ,尧德中 1,2
1,2
1,2
1 电子科技大学成都脑科学研究院临床医院,神经信息教育部重点实验室,生命科学与技术学院,成都
市,611731
2 电子科技大学中国—古巴神经技术与脑器交互“一带一路”联合实验室,成都市,611731
【摘 要】 类脑人工神经网络(brain-inspired artificial neural networks)作为当前人工智能领域的研究热点,其建模
体系已逐渐形成多维度发展的格局。该文系统梳理了类脑人工神经网络建模的研究进展,并在一个四维框
架下展开综述:①结构建模,主要关注神经元层面建模及网络拓扑优化;②功能建模,重点阐述注意力、
记忆、认知与情绪等机制的抽象实现方式;③结构–功能耦合,探讨结构与功能之间的紧耦合建模路径;
④类脑学习机制,聚焦多种生物启发的学习规则及其行为表现,特别以线虫、猕猴等经典神经科学模式生
物以及其他特定生物神经系统为例,总结了其对类脑人工神经网络设计与能效优化的启示,这也为系统比
较不同类脑模型并整合多种机制提供了更清晰的分析视角。最后,该文从不同医学影像的核心任务出发,
总结了类脑人工神经网络模型在医学影像分析中的创新应用,探讨了其在时空模式建模方面的潜力,并指
出多模态融合等关键挑战。未来研究将致力于推动结构、功能与学习机制的深度融合,并进一步拓展类脑
智能在临床等实际场景的广泛应用。
【关 键 词】 类脑人工神经网络;多维度建模;医学影像分析
【中图分类号】 TP391; R318
【文献标志码】 A doi: 10.12455/j.issn.1671-7104.250805
Multi-Dimensional Modeling of Brain-Inspired Artificial Neural
Networks and Its Application in Medical Image Analysis
1,2
1,2
1,2
【 Authors 】 WANG Caiwei , CHEN Silin , JIANG Xi , YAO Dezhong 1,2
1 The Clinical Hospital of Chengdu Brain Science Institute, MOE Key Laboratory for NeuroInformation,
School of Life Science and Technology, University of Electronic Science and Technology of China,
Chengdu, 611731
2 China-Cuba Belt and Road Joint Laboratory on Neurotechnology and Brain-Apparatus Communication,
University of Electronic Science and Technology of China, Chengdu, 611731
【 Abstract 】 Brain-inspired Artificial Neural Networks (ANNs) have become a prominent research focus in the field of
artificial intelligence, and their modeling framework has gradually evolved into a multidimensional
development paradigm. This article systematically reviews the progress in this field and organizes it within
a four-dimensional framework: ①structural modeling, which primarily focuses on neuron-level modeling
and network topology optimization; ②functional modeling, emphasizing the abstract implementation of
mechanisms such as attention, memory, cognition, and emotion; ③structure-function coupling, exploring
tightly integrated modeling pathways between structure and function; ④brain-inspired learning
mechanisms, concentrating on various biologically inspired learning rules and their behavioral
manifestations. Special attention is given to insights derived from classical neuroscience model
organisms, such as Caenorhabditis elegans, macaques, and other specific biological nervous systems,
which contribute to the architectural design and energy efficiency optimization of brain-inspired ANNs.
收稿日期:2025-11-17
基金项目:科技创新2030–重大项目(2021ZD0200800);国家自然科学基金(62576077,62276050)
作者简介:王采薇,E-mail: 2023140903016@std.uestc.edu.cn
通信作者:蒋希,E-mail: xijiang@uestc.edu.cn;尧德中,E-mail: dyao@uestc.edu.cn
167

