EARLY DETECTION OF COGNITIVE DECLINE USING VOICE ASSISTANT COMMANDS

Area of Science:

• Computational neuroscience
• Gerontology
• Speech processing

Background:

• Early detection of Alzheimer's Disease and Related Dementias (ADRD) is crucial for effective treatment and disease management.
• Machine learning models trained on spontaneous speech have shown promise in detecting and classifying ADRD.
• Voice-Assistant Systems (VAS) offer a novel platform for continuous monitoring and data collection from at-risk populations.

Purpose of the Study:

• To investigate the feasibility of using Voice-Assistant Systems (VAS) data for cognitive status classification.
• To develop unique feature sets from VAS interactions for machine learning model training.
• To assess the accuracy of machine learning models in classifying different stages of cognitive decline using VAS data.

Main Methods:

• Collected speech data from older adults with varying cognitive statuses (Dementia, Mild Cognitive Impairment, Healthy Control) interacting with VAS.
• Developed novel feature sets from the collected VAS data.
• Employed multi-class classification, binary classification, and regression models using the developed features.

Main Results:

• VAS data successfully classified participants into Dementia (DM), Mild Cognitive Impairment (MCI), and Healthy Control (HC) categories with up to 74.7% accuracy.
• The models achieved up to 62.8% accuracy in distinguishing between Healthy Control (HC) and Mild Cognitive Impairment (MCI) participants.
• Feature sets derived from VAS interactions proved effective for cognitive status classification.

Conclusions:

• Voice-Assistant Systems (VAS) provide a viable and accurate method for collecting speech data for cognitive decline assessment.
• Machine learning models utilizing VAS data can significantly aid in the early detection and classification of ADRD.
• This approach holds potential for remote, non-invasive cognitive monitoring of at-risk individuals.