AI-Enabled Shoe Insoles to Assess Ambulatory Function in Spinal Muscular Atrophy and Duchenne Muscular Dystrophy

Topic:

Other

Poster Number: 157

Author(s):

Jacqueline Montes, PT, EdD, Columbia University Irving Medical Center, New York, NY, USA, David Uher, MS, Department of Rehabilitation and Regenerative Medicine, Columbia University Irving Medical Center, Ton Duong, MEng, Stevens Institute of Technology, Sally Dunaway Young, PT, DPT, Stanford University, Tina Duong, PhD, Stanford University, Rabia Farooquee, MD, BA, Stanford University, Abby Druffner, BS, Boston Children's Hospital, Amy Pasternak, PT, DPT, PCS, Boston Children’s Hospital, Harvard Medical School, Boston, USA, Maria Pinkham, PT, DPT, DSc, Boston Children's Hospital, Damiano Zanotto, PhD, Stevens Institute of Technology

Background:
Disease-modifying therapies for spinal muscular atrophy (SMA) and Duchenne muscular dystrophy (DMD) are promising, however, symptoms of weakness and impaired function persists. Improvements are reported by patients but elude standard in-clinic examination. A compelling need to develop sensitive, quantitative assessments in real-life settings exists. Current wearable systems have limited use due to their modest accuracy in measuring spatiotemporal gait parameters. The purpose of this study was to validate spatiotemporal and kinetic gait parameters collected with novel instrumented insoles and new learning-based abstraction models.

Methods:
Ambulatory individuals with SMA/DMD and healthy controls (HC), ≥5 years, participated. Assessments included 6-minute walk test and simulated real-life tasks including curve walking, turns, and gait initiation and termination. Spatiotemporal and center of pressure (COP) gait parameters were collected with instrumented insoles (AI-Sole). A validated electronic walkway (Protokinetics-Zeno) provided reference data to estimate measurement error. Accuracy of AI-Sole was evaluated using Mean Absolute Error (MAE) and standard deviation of the error (SD) for stride length (SL), stride velocity (SV), stride time (ST), swing time (SwT), stance time (StT) and for the anteroposterior (AP) and mediolateral (ML) COP projections.

Results:
Twenty-two SMA/DMD and 13 HC (5.2-63.9 years) participated. MAE for spatiotemporal and COP parameters were similar between SMA/DMD and HC. In SMA/DMD MAE(SD) was SL: 4.10(1.01) cm, SV: 4.02(1.38) cm/s, ST: 0.013(0.009) s, SwT: 0.032(0.020) s, and StT: 0.033(0.021)s. For COP, MAE(SD) was 0.36(0.36) cm and 0.88(0.96) cm for ML and AP projections, respectively.

Conclusions:
AI-Sole accurately assessed spatiotemporal parameters and COP trajectories. Measurement error was low when compared to the reference system. Emerging wearable devices allow for ubiquitous monitoring of mobility but use is hampered by moderate accuracy. Foot-worn devices hold promise for real-world gait analysis since they are minimally obtrusive, less susceptible to sensor drift than multi-sensor systems, and offer better accuracy and granularity than wrist-worn conventional sensors.

Acknowledgements: Funding was provided by the Muscular Dystrophy Association (MDA629259) and in part by Cure SMA.