Combining Continuous Smartphone Native Sensors Data Capture and Unsupervised Data Mining Techniques for Behavioral...

Area of Science:

• Digital Health
• Psychiatry
• Machine Learning

Background:

• Smartphones and wearable sensors enable nonintrusive collection of daily activity data.
• This data can capture health-related events like mobility patterns without active patient participation.
• A system was developed to detect mobility pattern changes using smartphone sensors and machine learning.

Purpose of the Study:

• To assess the feasibility of detecting mobility pattern changes in outpatients with depression using smartphone sensors.
• To evaluate an unsupervised detection technique for analyzing smartphone-acquired data.

Main Methods:

• 38 outpatients with depression participated from Hospital Fundación Jiménez Díaz.
• The Evidence-Based Behavior (eB2) app collected data including inertial sensors, physical activity, call/message logs, app usage, Bluetooth, Wi-Fi, and location.
• A change-point detection technique was applied to location data from 9 outpatients.

Main Results:

• The unsupervised detection technique identified specific mobility pattern changes based on patient activity.
• These detected changes may serve as indicators of behavioral and clinical state shifts.
• Results from 5 patients are presented as a case series.

Conclusions:

• The developed technique can automatically detect mobility pattern changes in outpatients.
• These detected mobility changes may indicate potential relapses or clinical shifts.
• Detected changes do not always correlate with relapses, and some clinical changes may not be detected.