The repository contains data and code notebooks for the study on generalizability of machine learning models for human activity recognition in Healthy Subjects and Parkinson's disease patients. The study objective is to train ML models on exercise data from healthy subjects to accurately identify and measure the intensity of physical activity in PD patients, which can help in personalized exercise recommendations. The models were trained on data from medically validated triple synchronous sensors connected to 8 non-PD subjects performing 32 exercises and tested on 8 PD patients at different stages of the disease. The results indicate that better data preprocessing before modeling can provide some model generalizability, but it is still challenging as the models work with high accuracy on only one group (healthy or PD patients) and not both.
This work has clinical relevance as it can help clinicians, caregivers, and apps in accurately measuring physical activity in PD patients and making personalized exercise recommendations to augment medications that reduce tremors and stiffness.
If you use this work, please cite:
Aswar, S., Yerrabandi, V., Moncy, M. M., Boda, S. R., Jones, J., & Purkayastha, S. Generalizability of Human Activity Recognition Machine Learning Models from non-Parkinson’s to Parkinson’s Disease Patients.