Background
Facioscapulohumeral muscular dystrophy (FSHD) is a progressive muscle dystrophy characterized by weakness and wasting of facial, shoulder girdle, upper arm, trunk and lower extremity muscles. Clinical severity is often scored using the Lamperti score or via functional testing. Outcome measures that accurately quantify disease progression and quality of life are urgently needed. In this study we explore the use of digital technologies (CHDR MORE and Withings Health) in FSHD to measure social and physical activity and to evaluate their correlation to the clinical scores.
Methods
Patients with FSHD (n=38) and non-FSHD controls (n=20) were monitored for six weeks. Lamperti and the Timed Up and Go (TUG) test were scored. Smartphone sensors (location, inertial measurement unit, microphone, and ambient light) and phone usage (e.g. app usage and number of calls and texts) were continuously recorded using the CHDR MORE application. Biometric data was collected using Withings Steel HR smartwatch (steps, sleep and heartrate), Body+ scale (weight, body composition), and Blood Pressure Monitor. Analyses were performed by linear regression and random forests using Python package sklearn.
Results
The regression models demonstrated correlation of a composite of sleep quality, app use, location and physical activity with Lamperti scores (R2: 0.54; MSE: 4.89) and a composite of sleep quality, app use, location, physical activity and texting behavior with TUG (R2: 0.85; MSE: 1.0). The random forest could classify FSHD patients and controls with 85% accuracy and 100% sensitivity using an aggregate of 22 physical and behavioral features.
Conclusion
We found that 17 features were correlated with FSHD symptom severity while 22 features were sufficient to accurately classify FSHD patients vs. controls. This data indicates that mobile phone data and biometrics can be used to quantify disease severity in FSHD patients and potentially to monitor disease progression.