基于语义特征识别的阿尔茨海默病早期筛查方法

Early Screening Method for Alzheimer’s Disease Based on Semantic Feature Recognition

摘要: 阿尔茨海默病因其不可逆转的特性，早期筛查和治疗显得尤为重要。然而，传统的磁共振脑成像等筛查方法成本较高，难以广泛应用。相较之下，通过语音信号进行识别不仅便捷，而且更易于实现。本文提出了一种基于声音的语义特征筛查阿尔茨海默病的多分类器融合方法，利用机器学习技术，通过对语义特征的识别，实现对早期阿尔茨海默病患者的检测。本文收集了阿尔茨海默病患者的声音数据，构建了一个包含语义特征、受试者性别、年龄及SCS量表等特征的数据集。采用LASSO算法进行特征筛选后，构建了LR、RF、SVM、XGBoost等多种机器学习模型，并通过多分类器融合的集成方法实现阿尔茨海默病患者与正常人的分类。实验结果显示，本文方法的准确率达到89%，优于同类文献，能够有效辅助阿尔茨海默病的筛查。

Abstract: Alzheimer’s disease, due to its irreversible nature, makes early screening and treatment particularly important. However, traditional screening methods such as magnetic resonance imaging are costly and difficult to apply widely. In contrast, recognizing through voice signals is not only convenient but also easier to implement. This paper proposes a multi-classifier fusion method based on sound semantic feature screening for Alzheimer’s disease, utilizing machine learning techniques to detect early-stage Alzheimer’s patients through semantic feature recognition. We collected voice data from Alzheimer’s patients and constructed a dataset that includes semantic features, subject gender, age, and SCS scale among other characteristics. After feature selection using the LASSO algorithm, we built various machine learning models including LR, RF, SVM, and XGBoost, and achieved classification between Alzheimer’s patients and healthy individuals through an ensemble method of multi-classifier fusion. Experimental results show that the accuracy of this method reaches 89%, surpassing similar literature, and can effectively assist in the screening of Alzheimer’s disease.