The effects of accelerometer sensor position on freezing gait ratio parameters

Abstract

Freezing of gait (FoG) is an episodic gait pattern characterised by the inability to step that occurs on initiation or turning while walking for those with Parkinson’s disease (PD) [1]. This phenomenon is one of the most disabling yet poorly understood symptoms. It has been shown that tasks requiring bilateral limb coordination are the most likely to elicit FoG in the laboratory. Among the most promising tasks are stepping in place [2], walking and turning [3], and turning in place[4]. Previously, the Freezing Ratio parameter (FoG-ratio) has been developed to objectively measure freezing severity[5]. Usually, a lower limb acceleration signal in an antero-posterior direction measured by an inertial sensor has served as the source for its calculation[6,7]. Growing interest in single sensor utilisation in gait analysis brings up the question of whether any sensor other than the foot can measure freezing severity via FoG-ratio. Is FoG-ratio computed from a sensor located on the sternum or lower back comparable to the foot FoG-ratio during a walking turn? We included 34 Parkinson disease patients (21 males, 13 females), mean age 59.0 (SD 12.3) years in the study. All subjects performed an instrumented extended Timed Up&Go Test (TUG) wearing six synchronised inertial measurement units (Opals, APDM, USA) fitted via elastic straps. Sensors were located at the sternum, lower back, both wrists and feet. The turn subtask was automatically extracted from each TUG measurement. The FoG-ratio was calculated from antero-posterior acceleration acquired by a right foot sensor, left foot sensor, sternum (S) sensor, and lumbar (L) sensor. Depending on turn direction (left or right), each foot was denoted as the inner foot (IF) and outer foot (OF). Thus, four FoG-ratios (FoG_S-ratio, FoG_L-ratio, FoG_IF-ratio, FoG_OF-ratio) were obtained for each subject. The Kolmogorov-Smirnov test rejected the null hypothesis, i.e. data was not normally distributed. The Friedman test was employed for comparison of FoG-ratios. Posthoc pairwise comparisons were performed by Wilcoxon signed rank test (alpha level set to 0.05). Next, the Spearman correlation coefficient was calculated for all FoG-ratio pairs. The Friedman test revealed that the FoG-ratios from different sensor locations are statistically different (p<0.001). Pairwise tests showed statistically significant differences between the FoG_S-ratio and FoG_L-ratio (p<0.001), the FoG_S-ratio and FoG_IF-ratio (p=0.006), the FoG_L-ratio and FoG_IF-ratio (p=0.001), and the FoG_L-ratio and FoG_OF-ratio (p=0.001). The correlation analysis detected no significant relationship, Fig. 1.Download : Download high-res image (232KB)Download : Download full-size image Taking into account the results of location comparisons and their mutual relationships, no sensor seems to be a suitable alternative to foot sensors for freezing ratio calculation. However, additional analyses need to be performed before rejecting the possibility of employing other sensors in the analysis of freezing of gait in PD.