In-Home Passive Measurements of Mobility and Sleep in FSHD Patients and their Relationship with Clinical Disease Severity Metrics



Poster Number: 105


Rumen Hristov , Hariharan Rahul PhD, Zachary Kabelac PhD, Vicky Chan , Maya Hatch PhD, Jay Han MD, Diego Cadavid MD, Michelle Mellion MD, Dina Katabi PhD


1. Emerald Innovations, Inc., 2. Emerald Innovations, Inc., 3. Emerald Innovations, Inc., 4. UC Irvine, School of Medicine, Department of Physical Medicine and Rehabilitation, 5. University of California Irvine, 6. UC Irvine, School of Medicine, Department of Physical Medicine and Rehabilitation, 7. Fulcrum Therapeutics, 8. Fulcrum Therapeutics, 9. Emerald Innovations, Inc.

Objective: Assess the effectiveness of in-home mobility and sleep monitoring in capturing FSHD disease severity using Emerald, a contactless home sensor and machine learning platform.

Background: FSHD clinical trials suffer from the lack of accurate low-patient burden metrics that track disease severity and correlate with clinical outcomes. To address this problem, we monitored patients passively in their homes using Emerald, a wireless device that sits in the background of the home like a Wi-Fi router, and analyzed surrounding wireless signals to infer patients’ mobility and sleep.

Methods: 10 FSHD patients (7 females and 3 males aged 31-52, with Clinical Severity Scale (CSS) 1.5-4) were observed in their homes 24/7 for 3 months using Emerald.

We collected: 1) eTUG, the Emerald measured timed up and go, when the patient naturally gets up from a sleeping position in bed and walks for two meters; 2) In-Home Gait speed; 3) Sleep Schedule Variability (SSV), the variability across days in the time at which the patient goes to sleep.

These passive in-home metrics were correlated with metrics from clinic visits: CSS, Timed up and go (TUG), FSHD TUG (optimized TUG that includes truncal and lower limb impairments), Neurology Upper and Lower Extremities (NeuroUE, NeuroLE), Reachable Work Space (RWS), and PROMIS.

Results: Emerald passively captured 535, 3797, and 707 measurements of eTUG, Gait Speed, and SSV respectively. The median eTUG demonstrates very strong correlation with CSS (ρ=0.88, p=0.001) and good correlation with in-clinic FSHD metrics: TUG (ρ=0.73, p=0.017), FSHD TUG (ρ=0.73, p=0.017), NeuroLE (ρ=-0.73, p=0.017), NeuroUE (ρ=0.69, p=0.027), and PROMIS (ρ=-0.66, p=0.038). In-home Gait Speed has good correlation with TUG (ρ=-0.73, p=0.018) and FSHD TUG (ρ=-0.68, p=0.032). SSV demonstrates very strong correlation with CSS (ρ=0.8, p=0.005) and good correlation with NeuroUE (ρ=0.72, p=0.018), PROMIS (ρ=-0.68, p=0.03) and RWS (ρ=-0.68, p=0.032). Further, the metrics are sensitive: The Minimum Detectable Change (MDC95) for eTUG and Gait Speed after 1000 measurements are 4.65% and 1.17% respectively.

Conclusions: The results show that in-home mobility and sleep monitoring using Emerald allows obtaining a large number of measurements while imposing no significant burden on patients, who just go about their lives. Further, it exhibits good to strong correlations with clinical FSHD measures, and can enable sensitive longitudinal studies.