Home-Based Measurements of Dystonia in Cerebral Palsy Using Smartphone-Coupled Inertial Sensor Technology and Machine...

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Accurate dystonia severity measurement is crucial for managing treatments in dyskinetic cerebral palsy (CP).
• Current visual assessment methods are subjective, time-consuming, and capture only intermittent data.
• Dystonia fluctuates, making real-time, objective monitoring challenging in clinical settings.

Purpose of the Study:

• To investigate the feasibility of using smartphone-coupled inertial sensors and machine learning for home-based dystonia assessment in pediatric CP.
• To develop and evaluate machine learning models for objective, real-world dystonia severity evaluation.

Main Methods:

• Collected video and inertial sensor data from 12 pediatric CP patients during home-based activities.
• Clinicians scored dystonia severity from videos using the Dyskinesia Impairment Scale.
• Trained and cross-validated machine learning models using coupled clinical scores and sensor data.

Main Results:

• Individually trained machine learning models achieved average F1 scores of 0.67 (lower extremities) and 0.68 (upper extremities) in detecting dystonia.
• A generalized model showed lower performance (F1 scores 0.45 and 0.34) but could distinguish high/low scores.
• Results indicate the potential for automated, objective dystonia detection using home-based sensor data.

Conclusions:

• Home-based smartphone sensor data combined with machine learning show promise for objective dystonia assessment in pediatric CP.
• This approach can provide frequent, real-world data to supplement traditional clinical evaluations.
• Future work requires more data to improve model generalizability and assess long-term performance.