Multiscale and Shannon entropies during gait as fall risk predictors—A prospective study

Publication date: Available online 9 November 2016
Source:Gait & Posture
Author(s): Lucia Bizovska, Zdenek Svoboda, Nicolas Vuillerme, Miroslav Janura
Although entropy-based measurements of gait dynamics are becoming widely used tools for fall risk assessment, their relationship to fall occurrence is still unclear. The aim of this study was hence to compare fallers and non-fallers in terms of gait dynamics assessed by the multiscale and Shannon entropy. This study included 139 participants, aged 60–80 years, divided into two groups according to fall occurrence during a 6-month prospective observation (38 fallers, 101 non-fallers). The methodology involved the use of the Tinetti balance assessment tool (TBAT) and 5minutes of overground walking with 3D accelerometers located near the L5 vertebra and shanks. We analyzed 150 strides for gait complexity, an index of complexity (CI), computed from multiscale entropy (MSE) and Shannon entropy (ShE) derived from the recurrence quantification analysis. We found no significant differences between groups in MSE and CI. The TBAT total score was significantly higher in non-fallers (P=0.033), however, both groups showed low risk of falls. ShE in the anterior-posterior direction from trunk and in the mediallateral direction from the shanks were both significantly higher in fallers (P=0.020; P=0.024). ShE was negatively correlated with CI, the shank ShE in the vertical direction was positively correlated with TBAT. Taken together, our findings suggest that MSE is not able to distinguish between highly functional groups, whereas Shannon entropy seems to be sufficient in fall risk prediction.



from #Audiology via ola Kala on Inoreader http://rss.sciencedirect.com/action/redirectFile?&zone=main&currentActivity=feed&usageType=outward&url=http%3A%2F%2Fwww.sciencedirect.com%2Fscience%3F_ob%3DGatewayURL%26_origin%3DIRSSSEARCH%26_method%3DcitationSearch%26_piikey%3DS0966636216306385%26_version%3D1%26md5%3Da16df30e50bb014e47eae7bb1b3406c4
via IFTTT