Accelerometric Classification of Resting and Postural Tremor Amplitude

Area of Science:

• Neurology
• Biomedical Engineering
• Wearable Technology

Background:

• Clinical tremor assessments suffer from low resolution and rater bias, limiting their use in outpatient settings and treatment evaluation.
• Objective, quantitative methods are needed to assess tremor severity and treatment efficacy, particularly for deep brain stimulation (DBS).

Purpose of the Study:

• To investigate accelerometry metrics for objectively classifying tremor amplitude in Essential Tremor (ET) and Parkinson's Disease (PD).
• To compare accelerometry-derived metrics with clinical ratings of tremor severity.

Main Methods:

• Collected triaxial accelerometry data from finger-worn and wrist-worn sensors during resting and postural tremor tasks in 16 ET and 25 PD patients.
• Utilized spectral components and pure acceleration metrics, applying nonlinear regression to relate sensor data to clinical ratings.
• Validated the model on a holdout dataset to assess predictive performance and interrater reliability.

Main Results:

• All tested accelerometry metrics demonstrated high concordance (>70%) in predicting clinical tremor ratings.
• Substantial interrater reliability (Cohen's weighted Kappa > 0.7) was achieved for the accelerometry-based predictions.
• Finger-worn accelerometry showed slightly superior performance compared to wrist-worn devices.

Conclusions:

• Triaxial accelerometry provides a reliable and objective method for quantifying resting and postural tremor amplitude in ET and PD patients.
• This technology offers a viable alternative to subjective clinical scales, especially for evaluating interventions like DBS.
• The release of the dataset and software facilitates further research and development in tremor assessment.