实用医学杂志 ›› 2025, Vol. 41 ›› Issue (9): 1413-1424.doi: 10.3969/j.issn.1006-5725.2025.09.021

• 综述 • 上一篇    

深度学习在阿尔茨海默病疾病转化预测影像学研究中的应用价值

韩英妹1,李一杰1,张衡1,李伟庆1,冯泽1,王丰2()   

  1. 1.黑龙江中医药大学研究生院 (黑龙江 哈尔滨 150040 )
    2.黑龙江中医药大学附属第一医院CT磁共振科 (黑龙江 哈尔滨 150040 )
  • 收稿日期:2024-12-27 出版日期:2025-05-10 发布日期:2025-05-20
  • 通讯作者: 王丰 E-mail:wfzmy123@163.com
  • 基金资助:
    国家自然科学基金面上项目(81973930);黑龙江省自然科学基金资助项目(LH2023H065);黑龙江中医药大学研究生创新科研项目(2024yjscx114)

The application value of deep learning in imaging studies for predicting the conversion of Alzheimer′s disease

Yingmei HAN1,Yijie LI1,Heng ZHANG1,Weiqing LI1,Ze FENG1,Feng WANG2()   

  1. Graduate School of Heilongjiang University of Chinese Medicine,Harbin 150040,Heilongjiang,China
  • Received:2024-12-27 Online:2025-05-10 Published:2025-05-20
  • Contact: Feng WANG E-mail:wfzmy123@163.com

摘要:

阿尔茨海默病(Alzheimer's disease, AD)作为一种神经退行性疾病,在无症状阶段脑部便已呈现病理改变。随病理负荷加剧,患者不仅在记忆、语言、空间力、执行力、计算力等多个认知领域出现功能退化,还可能伴有情绪异常。一旦进展为AD,治疗难度极大。因此,早期诊断和精确预测疾病转化成为AD防治的核心任务,也是当前亟待攻克的科研难题。深度学习(deep learning, DL)模型在AD疾病诊断、预测、分类及特征提取等方面展现出不容忽视的优势,为解决这一难题带来新的希望。本研究首先简要介绍了AD疾病的转归和深度学习的基本知识,之后对深度学习在预测疾病转化的影像学研究从两个方面进行概述:一方面总结目前有关AD疾病分类和预测性能创新的深度学习模型;另一方面概述应用于AD诊断、分类和预测方面的深度学习融合模型。最后,本文阐述了此领域研究即将面临的挑战。本文表明深度学习模型是AD研究领域较为前沿的手段。

关键词: 深度学习, 卷积神经网络, 阿尔茨海默病, 磁共振成像

Abstract:

Alzheimer's disease (AD), a neurodegenerative disorder, manifests pathological changes in the brain even during the asymptomatic stage. As the pathological burden intensifies, patients experience functional decline in multiple cognitive domains, including memory, language, spatial perception, executive function, and calculation, and may also exhibit emotional abnormalities. Once AD progresses, treatment becomes extremely challenging. Therefore, early diagnosis and accurate prediction of disease conversion are core tasks in the prevention and treatment of AD, and they are also urgent scientific research challenges to be overcome. Deep learning (DL) models demonstrate considerable advantages in the diagnosis, prediction, classification, and feature extraction of AD, offering new hope for solving this challenging problem. This research commences with a concise introduction to the outcomes of AD and the fundamental knowledge of deep learning. Subsequently, it offers an overview of the imaging studies on the utilization of deep learning for predicting disease transformation from two perspectives. Firstly, it systematically summarizes the existing DL models that have demonstrated innovation in the classification and prediction performance of AD. Secondly, it provides a comprehensive outline of the DL fusion models applied to the diagnosis, classification, and prediction of AD. Finally, this paper expounds upon the impending challenges in the research of this domain. This article demonstrates that deep learning models is cutting-edge trends in the exploration of AD research.

Key words: deep learning, convolutional neural network, Alzheimer's disease, magnetic resonance imaging

中图分类号: