A Multimodal Approach for Early Alzheimers Disease Detection Using Handwriting and Speech Analysis

Abstract

The early detection of Alzheimer's Disease (AD) is important to slow the disease's progression and manage patient care. But conventional diagnosis techniques are often expensive, invasive, and not scalable for screening. The proposed solution integrates handwriting recognition and speech recognition for early diagnosis of AD in a non-invasive way. The approach employs a Random Forest model for handwriting features and a Convolutional Neural Network (CNN) for speech signals. The two modalities are integrated by a decision-level fusion scheme. The findings indicate that the developed model has an overall accuracy of 96.8%, surpassing the accuracy of the individual models, and has good robustness for practical screening.