Speech as a Window to the Brain: Innovative Techniques in Early Alzheimer’s Disease Detection

As dementia cases are projected to reach 150 million by 2050, with Alzheimer’s disease (AD) accounting for $60-70 \%$ of instances, early detection becomes crucial. AD and other dementias currently rank as the seventh leading cause of death globally. Our study addresses this by employing speech analysis for AD detection, offering a non-invasive and cost-effective diagnostic tool. Utilizing the ADReSS dataset, we integrated advanced Deep Learning techniques, including YAMNet and VGGish, with Mel Frequency Cepstral Coefficients (MFCCs) for feature extraction. An autoencoder was employed for dimensionality reduction to isolate key speech features. We applied Random Forest (RF), Artificial Neural Networks (ANN), and Support Vector Machines (SVM) for classification. The ANN model outperformed others, achieving an impressive $89.3 \%$ accuracy, surpassing current methodologies. This research highlights the efficacy of using speech analysis and ML, DL for early AD detection. Our approach, accessible and precise, promises advancements in early diagnosis and care for AD. Future work will aim to enhance model applicability to various stages of AD and related disorders.