Back to search

University of Leeds

Computer vision of video to measure bradykinesia and tremor in Parkinson’s Disease

Abstract

dc:description.abstract

The assessment of Parkinson’s disease is based upon clinician visual judgement. At the centre of this is a characteristic visible impairment of movement – bradykinesia – with often another visible sign, tremor. My thesis is that computer interpretation of video can provide clinically meaningful measures of finger tapping bradykinesia, and hand tremor, in Parkinson’s disease. A scoping literature review of technologies to automate the finger tapping test for bradykinesia in Parkinson’s (to 2021) identified 54 studies. Published methods include surface contact, infrared, gyroscope, accelerometer. There is a wide variation in strength and significance of correlations with clinical ratings, classification accuracies and group mean differences. Interrater reliability for judging finger tapping bradykinesia was investigated for 21 neurologists using 137 videos rated by the Movement Disorder Society revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS). There was only moderate agreement, intraclass correlation coefficient 0.53 (standard linear model) and 0.65 (cumulative linked mixed model). 24% of control videos were judged as bradykinesia. 70% of videos were correctly identified as Parkinson’s/control. A computer vision optical flow method was applied to 70 finger tapping videos, with dimensionality reduction using principal component analysis before input to classification models. Test accuracy was 0.8 for mild/moderate/severe bradykinesia and 0.67 for the presence of Parkinson’s disease. The computer vision pose estimation technique DeepLabCut was applied to 133 finger tapping videos. Resultant measures correlated well with clinical ratings of bradykinesia (Spearman coefficients): −0.74 speed, 0.66 amplitude, −0.65 rhythm for Modified Bradykinesia Rating Scale; −0.56 speed, 0.61 amplitude, −0.50 rhythm, −0.69 combined for MDS-UPDRS. All p < .001. Eulerian video magnification was applied to 48 videos of atremulous hands. The proportion of hands correctly classified as parkinsonian/control by clinicians was higher after Eulerian magnification (OR = 2.67; CI = [1.39, 5.17]; p < 0.003). Optical flow with Fourier transform was applied to 40 videos of tremulous hands. Bland-Altman analysis of dominant tremor frequency from video compared with accelerometer showed excellent agreement: 95% limits of agreement −0.38 Hz to +0.35 Hz. These results suggest that standard smartphone video can be used to derive measures of bradykinesia and tremor, and could form the basis of a tool to augment clinical assessment.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Leeds
Year dc:date.issued
2022

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Williams, Stefan
Advisors dc:contributor.advisor
  • Relton, Samuel D
  • O'Connor, Rory
  • Alty, Jane E

Rights

Language dc:language.dcterms_iso639_2
eng

Chain of custody

source
Harvested from
White Rose University Consortium
Base URL
etheses.whiterose.ac.uk/cgi/oai2
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Williams, Stefan. Computer vision of video to measure bradykinesia and tremor in Parkinson’s Disease. doctoral thesis, University of Leeds, 2022.