Activity-aware essential tremor evaluation using deep learning method based on acceleration data.

Area of Science:

• Neurology
• Biomedical Engineering
• Computer Science

Background:

• Essential tremor (ET) is a common neurological disorder.
• Current ET assessment relies on subjective clinical rating scales.
• These scales may underestimate tremor fluctuations during daily activities.

Purpose of the Study:

• To develop an automated system for rating ET severity using motion sensors and deep learning.
• To accurately assess ET during voluntary human activities.
• To propose a blockchain solution for anonymous tremor data sharing.

Main Methods:

• A smartwatch-based system collected motion data from 20 ET subjects during standard tasks.
• Deep learning models were developed for activity classification (ACMs) and tremor evaluation (TEMs).
• Tremor severity was rated by two neurologists using the Fahn-Tolosa Marin Tremor Rating Scale (FTMTRS).

Main Results:

• Activity classification models achieved high accuracy (89.73%-98.84%) for six activities.
• Tremor evaluation models showed significant correlation with neurologist FTMTRS ratings (r=0.92-0.93).

Conclusions:

• Motion sensor data combined with deep learning can effectively classify human activities.
• This approach provides an objective method for evaluating ET severity during various activities.