Alzheimer's Disease Detection and Dataset Creation from Spontaneous Speech

Abstract

This study presents a novel approach for the diagnosis and prediction of the stages of Alzheimer's disease. The aim of the study is to create a dataset for Alzheimer's disease based on Turkish speech data and to provide a basis for future studies. For this purpose, spontaneous speech of Alzheimer's patients was analyzed and language and speech disorders were analyzed. The study detects signs of Alzheimer's disease based on the patient's speech data instead of clinical test results or imaging data. Using machine learning and deep learning techniques, language features were analyzed and the results of different methods were compared. According to the results, 64% success rates were obtained with SVM and 70% with BERT+RNN. These results show that language-based analysis can be effective in early diagnosis and detection of Alzheimer's disease. Furthermore, this study provides a valuable resource for studying Alzheimer's disease on Turkish speech data and contributes to the development of language-based diagnosis methods.