Fatigue Detection with Machine Learning Approaches using Data from Wearable Devices.

Area of Science:

• Biomedical Engineering
• Digital Health
• Wearable Technology

Background:

• Chronic fatigue is a primary symptom in immune-mediated inflammatory diseases like Systemic Lupus Erythematosus (SLE) and Sjögren's disease (SjD).
• Current fatigue assessments primarily use subjective self-report questionnaires.
• Objective, passive measures are needed to complement subjective data and offer deeper insights into fatigue's impact on daily life.

Purpose of the Study:

• To objectively estimate fatigue in individuals with SLE and SjD using accelerometer data and machine learning.
• To compare objective fatigue measures against demographically matched healthy volunteers (HNV).
• To explore the potential of wearable devices for developing digital biomarkers of fatigue.

Main Methods:

• Collected accelerometer data from 96 participants over 24 weeks using ActiGraph Centrepoint Insight Watch in a free-living setting.
• Extracted activity-based, sleep-based, and circadian rhythm features from raw accelerometer data.
• Trained a machine learning classifier to distinguish daily fatigue status (fatigued/not fatigued) and evaluated using cross-validation.

Main Results:

• The machine learning model effectively distinguished fatigue status with better than random performance.
• Accelerometer features alone performed similarly to baseline participant characteristics alone in fatigue detection.
• Model performance (ROC-AUC) ranged from 0.44–0.70 in individual cohorts, improving to 0.76–0.83 when combined.

Conclusions:

• Wearable devices show significant potential for developing objective digital biomarkers of fatigue.
• These digital biomarkers could aid in assessing treatment response across various therapeutic areas.
• Objective fatigue monitoring can provide valuable complementary data to subjective patient reports.