A Novel Speech-Aware Framework (SACENet) for Non-Invasive Detection of Alzheimer’s Disease

Language, memory, and cognition are all impacted by Alzheimer's Disease (AD), a neurological disorder. Early detection is essential for timely intervention, but current diagnostic methods are often expensive, invasive, and non-scalable. The study suggests a deep learning-based technique called the Speech-Aware Cognitive Evaluation Network (SACENet) for the non-invasive prediction of Alzheimer's disease using speech data. SACENet combines a Convolutional Neural Network (CNN) for acoustic feature extraction with a Bidirectional Long Short-Term Memory (BiLSTM) network to model temporal speech characteristics such as articulation rate, pause frequency, and prosodic variation. The model was trained and evaluated using the ADReSS Challenge dataset, which includes balanced voice recordings from 78 AD patients and 78 healthy controls. SACENet's binary classification results included an F1-score of 89.28%, accuracy of 92.3%, precision of 90.12%, and recall of 88.46% after preprocessing steps like MFCC extraction and silence removal. These results demonstrate the effectiveness of the proposed method for early and precise speech-based AD detection, and it holds great potential for integration with remote diagnostic systems.