Automated Assessment Tests with Depth Sensors in Older Adults

Area of Science:

• Gerontology
• Biomedical Engineering
• Computer Science

Background:

• Mobility assessments are crucial for evaluating frailty and fall risk in older adults.
• Early detection of mobility changes is key to preventing severe health events.
• Current assessment methods can be time-consuming and subjective.

Purpose of the Study:

• To develop and validate an automated system for mobility assessments using depth imaging.
• To accurately predict Short Physical Performance Battery (SPPB) scores and Timed-Up-and-Go (TUG) test times.
• To classify older adults into frailty risk categories (high-risk, borderline, not-at-risk).

Main Methods:

• Utilized depth sensors to capture older adults performing SPPB and TUG tests.
• Developed a method to identify activities based on skeleton joints extracted from depth images.
• Employed an LSTM model for automated TUG timing prediction and SPPB score prediction.

Main Results:

• The automated SPPB system achieved 98% Mean Squared Error (MSE) accuracy in predicting SPPB scores.
• The automated TUG timing prediction achieved 94% MSE accuracy.
• The system successfully classified older adults into different frailty risk levels based on predicted scores.

Conclusions:

• Automated mobility assessments using depth imaging are highly accurate and reliable.
• This technology offers a promising tool for objective and efficient frailty assessment in clinical settings.
• Early identification of frailty risk can lead to timely interventions, improving health outcomes for older adults.