Objective: To investigate the predictive value of different accelerometry metrics for clinical ratings of Essential Tremor (ET) and tremor in Parkinson’s Disease (PD).

Background: To assess treatment effects of therapies like deep brain stimulation (DBS) on tremor, clinical rating scales are routinely used. Such scales have several limitations as a low 5‑point resolution and high rater-dependency and do not allow for continuous tremor assessment in ambulatory settings. To overcome these limitations, accelerometry can be a feasible and reliable technique to accurately measure tremor.

Method: We conducted tremor analysis in 16 ET and 25 PD patients in different DBS settings, resulting in a dataset of 860 trials of each resting and postural tremor in the upper limb. Tremor amplitude was clinically rated based on the respective MDS-UPDRS items and simultaneously recorded by two triaxial accelerometers (index finger and wrist). Spectral metrics such as i) area under the power spectral density curve, ii) peak power between 3 and 12 Hz, and acceleration metrics such as iii) mean acceleration, and iv) mean envelope of acceleration within these frequencies were calculated. The dataset was randomly split into a training dataset of 30 upper limbs (ET:PD ratio = 1:1, n = 479 trials) and a test dataset (n = 381). To determine the relationship between accelerometry and clinical ratings, nonlinear regression analysis was performed for each approach in the training dataset and then validated in the test dataset.

Results: Accelerometry metrics predicted clinical ratings with a high concordance (> 70 %) and substantial interrater reliability (mean Cohen’s weighted kappa κ_w > 0.7) in out-of-sample data. Acceleration-based metrics performed slightly better than frequency-based metrics with mean acceleration performing best for resting tremor (finger: κ_w = 0.795; wrist: κ_w = 0.771) and postural tremor on the finger (κ_w = 0.774). For postural tremor measured on the wrist the mean envelope metric performed best with κ_w = 0.714. Weighted kappa and concordance were slightly but consistently higher for finger than for wrist accelerometry.

Conclusion: In a large dataset and using out-of-sample validation, we demonstrate that accelerometry reliably predicts clinical ratings of tremor while offering multiple advantages over traditional rating scales.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/accelerometric-classification-of-resting-and-postural-tremor-in-essential-tremor-and-parkinsons-disease/