Speech and Language Biomarkers for Early Alzheimer's Detection: A Comparative Study Using SVM, Random Forest, and LSTM

This Early detection of Alzheimer's disease is crucial for timely intervention and effective management of cognitive decline. This study investigates the use of speech and language biomarkers combined with machine learning and deep learning algorithms SVM, Random Forest, and LSTM to detect early signs of Alzheimer's disease. Features such as speech rate, pause frequency, lexical diversity, syntactic complexity, and cognitive test scores were extracted and analyzed from patient data. Among the algorithms, SVM achieved moderate performance, Random Forest improved classification through ensemble learning, and LSTM demonstrated the highest accuracy of 97%, effectively capturing temporal and sequential patterns in speech. Comparative evaluation with prior studies shows that the proposed models outperform previously reported results, highlighting the potential of LSTM for accurate and noninvasive early Alzheimer's detection. The findings indicate that integrating linguistic and acoustic biomarkers with optimized machine learning and deep learning techniques provides a promising approach for reliable and efficient early diagnosis, offering significant implications for patient care and clinical decision-making.