Statistical Analysis of Dual-task Gait Characteristics for Cognitive Score Estimation

Area of Science:

• Gerontology
• Neurology
• Biomedical Engineering

Background:

• Traditional cognitive screening tools like the Mini-Mental State Examination (MMSE) are time-consuming and not suitable for frequent use.
• Dementia is known to impair patients' ability to perform multi-tasking efficiently.
• Gait analysis offers a potential objective measure for cognitive function assessment.

Purpose of the Study:

• To develop and validate an automated system for cognitive function screening using Kinect-based gait analysis.
• To investigate the effectiveness of single-task and dual-task performance in estimating dementia severity.
• To identify key gait characteristics crucial for automatic dementia score estimation.

Main Methods:

• An automated system using a Kinect device was implemented in three elderly facilities.
• Collected 10,833 behavior data points from 90 subjects performing locomotion and calculus tasks (single- and dual-task).
• Extracted 12 gait features and applied a machine learning algorithm for automatic dementia score estimation.

Main Results:

• The machine learning algorithm successfully classified subjects with an MMSE score of 23 or lower with 75.3% recall and 79.9% specificity.
• Gait characteristics were identified as important features for cognitive score estimation.
• Combining single- and dual-task gait features improved classification accuracy.

Conclusions:

• The automated Kinect-based system shows promise for objective and frequent cognitive function screening.
• Gait analysis, particularly under dual-task conditions, can effectively aid in dementia detection.
• This approach offers a non-invasive and potentially more efficient alternative to traditional cognitive tests.