Detection and Characterization of Walking Bouts Using a Single Wrist-Worn Accelerometer in Free-living Conditions.

Area of Science:

• Biomedical Engineering
• Neurology
• Wearable Technology

Background:

• Early detection of Parkinson's disease (PD) is crucial for effective management and treatment.
• Current methods often rely on invasive sensors in clinical settings, limiting real-world applicability.
• Actigraphy offers a non-invasive approach for long-term, free-living data acquisition.

Approach:

• Developed an algorithm using triaxial accelerometer data from actigraphy to detect walking bouts (≥10 seconds).
• Characterized walking bouts using cadence and arm swing parameters.
• Utilized a Walking Score (WS) based on the auto-correlation function (ACF) of the acceleration vector.

Key Points:

• Achieved high accuracy in walking bout detection: precision 0.90, recall 0.77, F1 score 0.83 compared to expert scoring.
• Successfully measured arm swing amplitude, a key gait parameter.
• Demonstrated the feasibility of using actigraphy for objective gait analysis in free-living conditions.

Conclusions:

• The proposed algorithm provides an accurate and non-invasive method for detecting gait abnormalities indicative of early Parkinson's disease.
• This approach can facilitate remote patient monitoring and early diagnosis in free-living environments.
• Further research can refine the algorithm for comprehensive PD progression monitoring.