摘要
针对现有阿尔茨海默病图像分类模型规模过大,难以被部署到移动端进行工作的问题,提出一种轻量化模型,该模型基于ResNet18模型作出如下改进:调整模型结构并使用深度可分离卷积替换部分普通卷积以降低模型的参数量;改进并使用空洞空间金字塔(ASPP)模块提升模型捕获全局上下文信息的能力;嵌入注意力模块CBAM提升模型特征提取效率。实验结果表明,相对于模型ResNet18,改进后的模型参数量减少95.51%,分类准确率、F1分数分别提高1.67百分点和2.20百分点,具有更高的应用价值。
Aimed at the problem that the existing Alzheimer's disease image classification model is too large to be deployed to the mobile terminal,a lightweight model is proposed.Based on the ResNet18 model,the following improvements were made:the model structure was adjusted and the depth separable convolution was used to replace part of the common convolution to reduce the number of parameters in the model;the void space pyramid(ASPP)module was improved to enhance the ability of the model to capture global context information.The embedding of attention module CBAM improved the efficiency of feature extraction.The experimental results show that compared with the model ResNet18,the number of parameters of the improved model is reduced by 95.51%,and the classification accuracy and F1 score are increased by 1.67 and 2.20 percentage points,respectively,which has higher application value.
作者
王青松
马金刚
宋佳航
李明
Wang Qingsong;Ma Jingang;Song Jiahang;Li Ming(College of Intelligence and Information Engineering,Shandong University of Traditional Chinese Medicine,Ji’nan 250355,Shandong,China)
出处
《计算机应用与软件》
北大核心
2025年第12期179-184,190,共7页
Computer Applications and Software
基金
国家自然科学基金面上项目(82174528)
山东中医药大学产教融合协同育人项目(CJ2021005,CJ2021003)。
