|
A 3D camera based automated T-GUGT tes This publication appears in: Parkinsonism and Related Disorders Authors: B. Jansen and R. Deklerck Publication Date: Feb. 2010
Abstract: Background and Aims: In the timed Get Up and Go Test (T-GUGT)to total time required to rise from a chair, walk 3m., turn, walkback and sit down again, is compared to a threshold. This test iswidely used in fall risk and mobility assessment. The total time ismeasured by an observer and hence prone to errors. The durationsof the different phases are typically not available. This small studyaims to investigate whether image processing methods applied to3D camera images could be applied to automate the T-GUGT test.Methods: Custom image processing, calibration and pose transitionmethods were developed and applied to 3D camera recordings of4 healthy controls, each performing 6 consecutive instances of theT-GUGT. The durations of the different phases of the test and thetotal time taken were compared using Pearson’s Correlation tothe respective durations obtained from manually annotating therecorded images at the frame level.Results: The calculated total duration of the test correlates wellto the ground truth (R2 = 0.761), while the correlation between thedurations of the individual phases ranges from 0.69 for the forwardwalking phase down to 0.2 for sitting down and rising from thechair.Conclusions: The initial results suggest that a camera basedautomatic T-GUGT test is feasible, but that more research followedby proper validation is required. The added value of the proposedapproach is that it additionally provides durations of the individualphases.
|
