Abstract Details

Title Gait Measures from Videos Detect Parkinsonian Gait in Older Adults with Dementia

Topic Movement Disorders

Presentation(s) S55 - Movement Disorders: Neuromodulation, Circuits, and Management (4:42 PM-4:54 PM)

Poster/Presentation
Number 007

Objective To assess the predictive power of gait parameters extracted from video for rating severity of parkinsonian gait on clinical scales in older adults with dementia.

Background Parkinsonian gait in older adults with dementia is commonly assessed on the gait criterion of the Unified Parkinson's Disease Rating Scale (UPDRS), and where it is felt to be medication-induced, by the Simpson-Angus Extrapyramidal Side Effects Scale (SAS). These assessments are administered by skilled clinicians and are thus completed cross-sectionally at large intervals. There is an opportunity to develop video-based gait evaluation for frequent and autonomous monitoring of parkinsonism in gait in residential environments.

Design/Methods A video camera installed on the ceiling of a hallway of a dementia care unit was used to record participants’ natural gait. 363 walking bouts from 14 participants were rated on the gait criterion of the UPDRS and SAS by an expert annotator. Gait parameters were calculated from 2D keypoints extracted using the OpenPose human pose estimation library. Nested ordinal logistic regression was used to evaluate if adding gait parameters to baseline patient features of age, sex, and fall history improved regression to UPDRS and SAS gait scores.

Results Adding cadence, number of steps in the walking bout, and the symmetry index and coefficient of variation of step time significantly improved regression to UPDRS gait when compared to a model using only baseline clinical features. The SAS gait model was significantly improved over the baseline by incorporating the gait features of cadence, average margin of stability, and symmetry index and coefficient of variation of step time.

Conclusions Gait parameters extracted from video capture valuable information that can be used to assess the severity of parkinsonism in natural gait. This suggests that automated monitoring of parkinsonism in gait can be achieved in non-clinical settings using non-obtrusive, consumer-grade camera systems.

Authors/Disclosures
Andrea Sabo PRESENTER	No disclosure on file
	No disclosure on file
	No disclosure on file
	No disclosure on file
	No disclosure on file

��ɫ��

��ɫ��