Continuous Monitoring of Physical Activity to Predict Friedreich Ataxia Symptoms

Topic:

Clinical Trials

Poster Number: S81

Author(s):

Ashkan Vaziri, PhD, Biosensics LLC, Ram kinker Mishra, PhD, BioSensics LLC, Newton, MA, USA, Adonay Sastre Nunes, PhD, BioSensics LLC, Newton, MA, USA, Jose Casado, MS, BioSensics LLC, Newton, MA, USA, McKenzie Wells, MS, Children's Hospital of Philadelphia, Philadelphia, PA, USA, David Lynch, MD, PhD, Children's Hospital of Philadelphia

Background:
FA is an autosomal recessive neurodegenerative disorder that gradually impairs coordination, balance, and mobility. Traditional clinical assessments of FA are subjective and may encounter challenges such as ceiling effects, daily fluctuations, and limited sensitivity to changes. Wearable sensors present a viable solution, enabling continuous monitoring of physical activity and mobility in real-world settings, offering valuable insights derived from daily living experiences.

Objective:
To develop a predictive model for Friedreich Ataxia (FA) symptoms using physical activity outcomes measured remotely through wearable sensors.

Methods:
Individuals diagnosed with FA wore a PAMSys™ pendant and two wrist sensors for a week to measure daily physical activity and goal-directed movements. Moreover, conventional FA outcome assessments, such as modified Friedreich’s Ataxia Rating Scale (mFARS), 9-hole peg tests, and activity of daily living (ADL) scales, were administered during their FA Clinical Outcome Measures Study examination. Feature selection was done through Spearman correlation analysis, and machine learning model performance was assessed using a leave-one-out cross-validation technique.

Results:
The study involved 40 participants diagnosed with FA (average age of 26.7 ± 1.5 years, including 19 females). We observed significant correlations between physical activity outcomes and clinical assessments, including total mFARS, mFARS Section E, mFARS Section B, and ADL scores, showing a moderate to high effect size. Machine learning models successfully accounted for the variance in the clinical scores, explaining between 21% and 51% of the variance within the test dataset.

Conclusions:
Our results provide a data-driven, objective, and remote monitoring approach to predict FA clinical outcomes, leading to better patient care and outcomes. Wearable technology allows for remote monitoring, reducing the need for frequent in-person clinical assessments. Patients can be monitored in their natural environments, leading to greater convenience and potentially lowering healthcare costs.