F61 Validation of a gait event detection algorithm during overground walking in Huntington's disease. (12th September 2022)

• Record Type: Journal Article
• Authors: Lozano-García, Manuel; Doheny, Emer; Mann, Elliot; Drew, Cheney; Busse-Morris, Monica; Lowery, Madeleine
• Is Part Of: Journal of neurology, neurosurgery and psychiatry. Volume 93(2022)Supplement 1
• Page Start: A58
• Page End: A58
• Publication Date: 2022-09-12
• Subjects: Teager-Kaiser energy -- Gait event detection -- Step time -- Step count -- Validation; Neurology -- Periodicals; Nervous system -- Surgery -- Periodicals; Psychiatry -- Periodicals; 616.8
• DOI: 10.1136/jnnp-2022-ehdn.152
• Languages: English
• ISSNs: 0022-3050

Abstract

Abstract : Background: Objective evaluation of gait impairment in Huntington's Disease (HD) is challenging both in clinical trials and clinical practice. Algorithms, such as Teager-Kaiser Gait Event Detection (TKGED), enable detection of initial (IC) and terminal (TC) contact, using accelerometers on the shanks [1]. TKGED has not been validated in HD despite known accuracy in healthy individuals. Aim: To assess the performance of TKGED using shank and thigh acceleration signals in participants with HD. Methods: Seventeen participants performed two 2-minute walking tests, wearing accelerometers on shanks (ActiGraph GTX9, 100Hz) and thighs (ActivPAL4, 40Hz). Video data were recorded as the criterion measure. To obtain IC and TC points, video data were manually annotated and TKGED was applied to accelerometer signals. Step counts and stride, stance and swing times were estimated for each sensor. Intraclass correlation coefficients (ICC) determined agreement between video and accelerometer step counts. Step count differences were assessed using coefficients of variation (CV) and signed rank tests. Kruskal-Wallis tests assessed differences between video and accelerometer stride, stance and swing times. Results: Excellent agreement was observed for step counts between video and both shank (ICC=0.97, CV=4.0%, p=0.21) and thigh (ICC=0.94, CV=5.3%, p=0.23) accelerometers ( Figure 1 ). Compared with video annotation, TKGED tended to underestimate stance time and overestimate swing time (p<0.001), but yielded accurate estimates of stride time (p=0.28). Conclusions: The TKGED algorithm accurately estimates step count and stride time in HD. Its potential for enhancing evaluation of HD gait parameters should be explored. [1] Flood et al, IEEE TBME 2020