Machine Learning Approaches for Dementia Detection Through Speech and Gait Analysis: A Systematic Literature Review

Area of Science:

• Neuroscience
• Computer Science
• Artificial Intelligence

Background:

• Dementia encompasses progressive neurodegenerative disorders like Alzheimer's disease.
• Early detection is critical for timely intervention and management.
• Artificial intelligence (AI) offers potential for machine learning (ML)-assisted early detection.

Purpose of the Study:

• To review state-of-the-art ML-based dementia prediction methods.
• Focus on non-invasive approaches like gait and speech analysis for cost-effective screening.
• Assess the potential of these methods for early cognitive health insights.

Main Methods:

• Systematic literature review following PRISMA protocol.
• Searched Scopus, Web of Science, and PubMed databases (2017-2022).
• Selected and reviewed 40 relevant studies.

Main Results:

• Support vector machines and deep learning were common ML methods.
• Multimodal approaches demonstrated superior prediction performance.
• Deep learning in gait analysis is emerging; whole-body movement features enhance accuracy.
• Speech analysis combining acoustic, linguistic, and cognitive features yielded better results.

Conclusions:

• ML, particularly non-invasive methods, shows significant potential for early dementia prediction.
• Comparable accuracies in manual and automatic speech analysis suggest future automated detection.
• Non-invasive gait and speech analysis are promising for cost-effective dementia screening.