基于人群特征的阿尔兹海默症分类方法  

Classification Method of Alzheimer's Disease Based on Population Characteristics

作　　者：胡建举 张晓龙[1] 曾燕[2] 胡斐斐 HU Jianju;ZHANG Xiaolong;ZENG Yan;HU Feifei(Department of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065;Brain Science and Advanced Technology Institute,Wuhan University of Science and Technology,Wuhan 430065)

机构地区：[1]武汉科技大学计算机科学与技术学院,武汉430065 [2]武汉科技大学脑科学先进技术研究院,武汉430065

出　　处：《计算机与数字工程》2024年第2期321-326,331,共7页Computer & Digital Engineering

基　　金：科技部重大项目(编号:2020AAA08600);国家重点研发计划(编号:2020YFC2006000)资助。

摘　　要：在阿尔兹海默症(Alzheimer’s disease,AD)分类研究中,图像、生物标志物等数据集包含的样本少,获取成本高。为应对这一问题,论文提出一种基于人群特征进行建模的方法,并在CMDS数据集上进行了实验。首先,使用PAR方法分析特征与AD之间的相关性,根据分析结果进行特征选择;然后,使用ADASYN算法解决训练集样本不平衡问题;最后,使用XGBoost算法进行训练,得到最终模型。该模型的准确率和召回率达到了79.5%和77.6%,AUC达到了0.83。实验结果证明了该方法的有效性。In Alzheimer's disease(Alzheimer's disease,AD)classification research,data sets such as images and biomarkers contain few samples and high acquisition costs.To cope with this problem,the paper proposes a method of modeling based on popu-lation characteristics,and conducts experiments on the CMDS data set.First,the PAR method is used to analyze the correlation be-tween features and AD,and select features based on the analysis results.Then,the ADASYN algorithm is used to solve the problem of unbalanced training set samples.Finally,the XGBoost algorithm for training is used to obtain the final model.The accuracy and recall rates of the model reached 79.5%and 77.6%,and the AUC reached 0.83.The experimental results prove the effectiveness of this method.

关 键 词：阿尔兹海默症 人群特征 ADASYN PAR 机器学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]