Related Experiment Video
Updated: Jun 13, 2026
07:25
An Objective and Child-friendly Assessment of Arm Function by Using a 3-D Sensor
Published on: February 12, 2018
6.9K
Using Wearable Sensors and Machine Learning to Assess Upper Limb Function in Huntington's Disease
Adonay S Nunes1, İlkay Yıldız Potter1, Ram Kinker Mishra1
1BioSensics LLC, 57 Chapel St, Newton, MA, USA.
Research Square
|June 17, 2024
Summary
Wearable sensors and machine learning can detect upper limb changes in Huntington's disease (HD) and prodromal HD (pHD). This technology aids early detection, remote monitoring, and treatment evaluation for neurological disorders.
Area of Science:
- Neurology
- Biomedical Engineering
- Data Science
Background:
- Huntington's disease (HD) impacts both upper and lower limb function, but clinical assessments primarily focus on lower limb symptoms.
- Wearable sensors offer a method to collect real-world data, complementing clinical evaluations and providing deeper insights into disease progression.
Purpose of the Study:
- To monitor upper limb function in individuals with HD, prodromal HD (pHD), and controls (CTR) using wrist-worn wearable sensors during daily activities.
- To utilize deep learning and machine learning models to analyze kinematic features of goal-directed movements (GDMs) and predict disease groups and clinical scores.
Main Methods:
- A wrist-worn wearable sensor (PAMSys ULM) was used to collect data over seven days from 16 HD patients, 7 pHD patients, and 16 controls.
- A deep learning model detected GDMs, and kinematic features were extracted. Statistical and machine learning models were employed for prediction tasks.
- High participant compliance (average 99%) ensured data quality for analysis.
Main Results:
- Significant differences in GDM features were observed between HD, pHD, and CTR groups.
- HD participants exhibited fewer long-duration GDMs compared to controls. The entropy of velocity zero-crossing length segments showed the largest effect size and strongest correlation with clinical scores.
- Classification models achieved 67% balanced accuracy for distinguishing groups, with 0.72 recall for the HD group. Regression models accurately predicted clinical scores, explaining up to 60% of the variance in the upper extremity function subdomain of the UHDRS.
Conclusions:
- Wearable sensors and machine learning show promise for early identification of phenoconversion in HD.
- This approach facilitates remote monitoring of HD patients and evaluation of treatment efficacy in clinical trials.
- The study highlights the potential of upper limb monitoring for comprehensive assessment and management of Huntington's disease.
