Reliability of Scoring Transforms of Wearable Sensor Signals in People with Parkinson’s Disease
Abdelwahab Elshourbagy1, Liran Ziegelman2, Timothy Harrigan3, Manuel Hernandez4, James Brasic5
1Misr University for Science and Technology, 2University of Illinois, 3Johns Hopkins University, 4Carle Illinois College of Medicine, University of Illinois at Urbana-Champaign, 5Bellevue Hospital Center
Objective:
To assess the reliability of FFT and CWT scoring in Parkinson’s disease.
Background:
Objective, scalable motor assessment is needed for the diagnosis and treatment of Parkinson’s disease (PD) and related conditions, especially in remote or resource-limited contexts. Transforming signals from wearable-sensor data from structured motor assessments may provide a low-cost complement to in-person clinical ratings, but agreement with in-person assessments for people with PD and healthy controls (HC) must be established.
Design/Methods:

Eighteen participants (11 PD, 7 HC) completed five standardized repetitive motor tasks while wearing triaxial accelerometers attached to the extremities (wrists and ankles) to capture movement signals. Raw signals were transformed into fast Fourier (FFT) and continuous wavelet (CWT) visual outputs. Trained raters, blinded to participant diagnosis and laterality, scored the transformed images of test and retest after one week. Agreement was evaluated using average-measure intraclass correlations (ICCs) from a two-way mixed-effects model with absolute agreement. Level differences between FFT and CWT ratings were assessed using Wilcoxon signed-rank tests on patient medians.

Results:

Excellent agreement was observed between blind ratings of signals and their transforms across groups and sessions: PD Test (ICC = 0.918), HC Test (ICC = 0.840), PD Retest (ICC = 0.928), HC Retest (ICC = 0.925). Wilcoxon tests showed no significant level differences between FFT and CWT: PD Test Z = −0.211, p = .833; HC Test Z = 0.000, p = 1.000; PD Retest Z = −0.879, p = .379; HC Retest Z = −1.897, p = .058.

Conclusions:

Blind FFT and CWT visual scoring showed strong internal agreement across raters and sessions, supporting the integration of transformed wearable data into remote, clinician-augmented PD monitoring. This approach can facilitate telehealth when clinic visits are limited by infectious diseases or conflicts and may serve as a biomarker framework for clinical trials of interventions in PD and related conditions.

 

 

10.1212/WNL.0000000000216900
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.