revoAD: Revolutionizing Alzhiemer’s Disease Diagnosis through Multimodal Machine Learning for Universal Screening via Speech and Handwriting Patterns

Alzheimer's Disease (AD) is a global public health concern that leads to cognitive decline and memory loss. Existing AD diagnosis methods are invasive, expensive, and time-consuming. Hence, a cost-effective, highly sensitive screening tool is imperative. This study employs machine learning (ML) to detect AD through speech and handwriting pattern analysis. Over 15,000 samples, including audio, handwriting, and cognitive data from AD patients and controls, were preprocessed with Mel-Frequency cepstral coefficient testing, image normalization, binarization, and feature extraction. Six ML models were trained to detect AD based on both speech and handwriting markers like slurred speech, abrupt sentence endings, pronounced forgetfulness, legibility, stroke information, and zone-based features, achieving a combined F1-Score of 96.2% using an 80/20 split. The "revoAD" mobile app, developed with React JavaScript and Python OpenCV, achieved a 97.6% training accuracy, 97.3% data validation accuracy, and 10x faster diagnosis, addressing healthcare disparities by offering low-cost screening, especially in underserved areas. This study leveraged machine learning for AD diagnosis, promising to improve early detection and healthcare access.