基于多示例学习与多尺度特征融合的阿尔茨海默病分类诊断模型  

Classification of Alzheimer’s disease based on multi-example learning and multi-scale feature fusion

作　　者：曾安[1] 帅志富 潘丹 林劲芝[3] ZENG An;SHUAI Zhifu;PAN Dan;LIN Jinzhi(School of Computer Science,Guangdong University of Technology,Guangzhou 510006,P.R.China;School of Electronics and Information Engineering,Guangdong University of Technology and Education,Guangzhou 510665,P.R.China;Department of Neurosurgery,The Second People’s Hospital of Guangdong Province,Guangzhou 510310,P.R.China)

机构地区：[1]广东工业大学计算机学院,广州510006 [2]广东技术师范大学电子与信息学院,广州510665 [3]广东省第二人民医院神经外科,广州510310

出　　处：《生物医学工程学杂志》2025年第1期132-139,共8页Journal of Biomedical Engineering

基　　金：国家自然科学基金项目(61976058);广州市科技计划项目(202103000034,202206010007,202002020090);广东省科技计划项目(2021A1515012300,2019A050510041,2021B0101220006)。

摘　　要：阿尔茨海默症(AD)分类模型通常会将整张大脑影像分割为体素块,并为之赋予与整张影像一致的标签,但并非每个体素块都与疾病密切相关。为此,本研究提出了基于弱监督多示例学习(MIL)和多尺度特征融合的AD辅助诊断框架,并从体素块内部、体素块之间以及高置信度体素块三个方面设计框架。首先利用三维卷积神经网络并融入多视角网络,提取体素块内部的深层次特征;再通过位置编码和注意力机制捕捉体素块间的空间关联信息;最后筛选高置信度体素块并结合多尺度信息融合策略,整合关键特征用于分类决策。模型分别在AD神经成像倡议(ADNI)数据集和开放获取系列成像研究(OASIS)数据集上进行性能评估。实验结果表明,所提框架在AD分类以及轻度认知障碍转化分类两项任务中,相较于其他主流框架,ACC及AUC分别平均提升了3%和4%,且可寻找到触发疾病的关键体素块,为AD辅助诊断提供了有效依据。Alzheimer’s disease(AD)classification models usually segment the entire brain image into voxel blocks and assign them labels consistent with the entire image,but not every voxel block is closely related to the disease.To this end,an AD auxiliary diagnosis framework based on weakly supervised multi-instance learning(MIL)and multiscale feature fusion is proposed,and the framework is designed from three aspects:within the voxel block,between voxel blocks,and high-confidence voxel blocks.First,a three-dimensional convolutional neural network was used to extract deep features within the voxel block;then the spatial correlation information between voxel blocks was captured through position encoding and attention mechanism;finally,high-confidence voxel blocks were selected and combined with multiscale information fusion strategy to integrate key features for classification decision.The performance of the model was evaluated on the Alzheimer’s Disease Neuroimaging Initiative(ADNI)and Open Access Series of Imaging Studies(OASIS)datasets.Experimental results showed that the proposed framework improved ACC and AUC by 3%and 4%on average compared with other mainstream frameworks in the two tasks of AD classification and mild cognitive impairment conversion classification,and could find the key voxel blocks that trigger the disease,providing an effective basis for AD auxiliary diagnosis.

关 键 词：阿尔茨海默症 体素块 多示例学习 多尺度 位置编码 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术] R749.16[医药卫生—神经病学与精神病学]