Gait Characterization in Duchenne Muscular Dystrophy (DMD) Using a Single-Sensor Accelerometer: Classical Machine...

Area of Science:

• Biomechanical analysis
• Pediatric neurology
• Digital health

Background:

• Gait deviations in Duchenne muscular dystrophy (DMD) are visually apparent but difficult to quantify outside specialized labs.
• Objective gait analysis in children with DMD typically requires laboratory equipment, limiting accessibility.
• Consumer-grade accelerometers offer a potential solution for accessible gait assessment.

Purpose of the Study:

• To quantify gait differences between children with DMD and typically developing (TD) peers using smartphone accelerometers.
• To apply machine learning (ML) to differentiate DMD from TD gait patterns.
• To assess the feasibility of using accessible technology for DMD gait monitoring.

Main Methods:

• Fifteen children with DMD and 15 TD children (ages 3-16) underwent walking/running tests with a waist-worn iPhone accelerometer.
• Gait data included speed-calibration tests, a 6-minute walk test, a 100m run, and free walking.
• Temporospatial gait features were extracted, and ML models were used to classify participants.

Main Results:

• Extracted gait features revealed reduced step length and increased mediolateral power in DMD children, consistent with known gait abnormalities.
• Machine learning models achieved up to 100% accuracy in differentiating DMD from TD children based on gait data.
• The study demonstrated the ability to capture DMD-associated gait characteristics across various speeds and age groups.

Conclusions:

• Smartphone accelerometer data, analyzed with ML, can effectively capture and differentiate DMD-associated gait characteristics.
• This approach offers a promising, accessible method for monitoring DMD progression and gait changes in children.
• Accessible digital health tools can significantly aid in the clinical assessment and management of pediatric neuromuscular disorders.