中南民族大学学报（自然科学版）2026，Vol.45Issue(2)：212-220,9.DOI:10.20056/j.cnki.ZNMDZK.20250843
融合多模态信息与位置编码的阿尔茨海默病诊断
Diagnosis of Alzheimer's disease via fusion of multimodal information and position encoding
摘要
Abstract
Alzheimer's Disease(AD),as a fatal neurodegenerative disease,holds immense significance for early diagnosis and precise prediction of pathological regions in delaying disease progression and improving patient prognosis.Although past research has made progress in automated diagnostic technologies,the interpretability of existing methods remains the most significant issue troubling clinical studies,despite their commendable diagnostic accuracy.Against this backdrop,a diagnostic model for Alzheimer's Disease is proposed that integrates three-dimensional position encoding with multimodal data.The model combines three-dimensional position encoding,Transformer self-attention mechanisms and Fully Convolutional Networks(FCN)to automatically extract effective features from three-dimensional medical imaging data,generating high-resolution disease probability maps representing the entire brain.Through a multimodal attention mechanism,the probability map is organically integrated with objective clinical information,achieving precise predictive diagnosis of AD while providing more interpretable aspects for the model's decision-making process.关键词
阿尔茨海默病/磁共振影像/全卷积网络/三维位置编码/多模态注意力Key words
Alzheimer's disease/magnetic resonance imaging/Fully Convolutional Network/three-dimensional position encoding/multimodal attention分类
信息技术与安全科学引用本文复制引用
刘蓉,刘汝璇,李广昶,柴新宇,谭桂梅,唐奇伶..融合多模态信息与位置编码的阿尔茨海默病诊断[J].中南民族大学学报（自然科学版）,2026,45(2):212-220,9.基金项目
湖北省重点研发计划资助项目(2022BAA037) （2022BAA037）
中央高校基本科研业务费专项资金资助项目(CZQ24015) （CZQ24015）
