Category: Technology
Objective: To identify which gait metrics better distinguish PD-OFF from PD Best-ON and PD from HC, using a dimensionality-reduction method
Background: Objective gait analysis using wearable sensors allows a better characterization of Parkinson’s Disease (PD) gait, with potential usefulness as a diagnostic and prognostic biomarker. The set of kinematic variables that better characterize PD gait and better differentiate PD from healthy controls (HC) has not been established yet.
Method: 17 PD (OFF and Best-ON) and 34 HC matched for age and sex were evaluated. For gait assessment, patients walked 3x along a 4-meter-long corridor at a self-selected speed wearing a full-body set of 15 inertial measurement units (IMUs). Kinematic data was collected to reconstruct each subject’s body motion using a 3D kinematic model of the skeletal system. Principal Component Analysis (PCA) was performed to reduce the number of kinematic variables, and the resulting Principal Components (PC) were used to train a tree-based classifier (DP OFF vs DP Best-On, DP OFF vs HC and DP Best-ON vs HC).
Results: Mean age and mean disease duration were 60.2±8.1 and 12.1±5.1 years, respectively. MDS-UPDRS III score was 50.4±11.3(OFF) and 21.8±8.9 (Best-ON). PD OFF and Best-ON were discriminated mainly by spaciotemporal and angular variables with 80% precision (80% sensibility and specificity). Spaciotemporal and nonlinear metrics differentiated PD-OFF from HC with 90% precision (100% sensibility and 80% specificity). Discrimination between PD Best-On and HC was possible with a 90% precision (100% sensibility and 80% specificity) using non-linear and asymmetry related metrics.
Conclusion: The capacity of levodopa to modulate spaciotemporal and angular metrics probably explains the ability of these metrics to distinguish between OFF and Best-On states. In the OFF state, fundamental differences between PD and HC fall on spatiotemporal metrics, whereas these metrics are not able to differentiate PD in the Best-On state from HC. Instead, this distinction is possible using metrics related to the dynamic stability of gait and asymmetry.
To cite this abstract in AMA style:
R. Barbosa, M. Medonça, R. Oliveira, M. Santos, A. Abreu, P. Bastos, P. Pita-Lobo, A. Valadas, L. Correia-Guedes, J. Ferreira, M. Rosa, R. Matias, M. Coelho. Gait analysis using Wearable Sensors: different gait metrics can differentiate Parkinson’s Disease Patients from Healthy Controls and Parkinson’s Disease patients in different medication status [abstract]. Mov Disord. 2022; 37 (suppl 2). https://www.mdsabstracts.org/abstract/gait-analysis-using-wearable-sensors-different-gait-metrics-can-differentiate-parkinsons-disease-patients-from-healthy-controls-and-parkinsons-disease-patients-in-different-medicati/. Accessed December 10, 2024.« Back to 2022 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/gait-analysis-using-wearable-sensors-different-gait-metrics-can-differentiate-parkinsons-disease-patients-from-healthy-controls-and-parkinsons-disease-patients-in-different-medicati/