Reliability of pressure insoles to assess normal ground reaction forces during gait of healthy people

Abstract

Objective Pressure insoles are developed to be placed in the shoes and provide embedded measurement of the pressure or/and normal ground reaction forces (NGrF). Recently, Pedoped® insoles were commercialized and consists in only one sensor providing NGrF in real-time biofeedback. As NGrF represents about 90% of ground reaction forces during gait, this device permits to record the lower limb loading during walking, standing or in other daily living activities. Thus, these insoles could be a useful tool in rehabilitation when dealing with gait asymmetry or load recovery. As the first step to evaluate the reliability of these insoles, the accuracy of NGrF measurement with Pedoped® insoles was assessed during gait of healthy subjects. Material/patients and methods Five healthy subjects were recruited. All subjects were wearing the same shoes with Pedoped® insoles placed inside. Data collection was carried out in a motion capture laboratory equipped with 4 AMTI® force platforms (PFF). Subjects were asked to stamp each foot on a PFF during one minute at self-selected speed. PFF and insoles data were synchronized by detecting the first NGrF above 20 N for the two systems. About sixty cycles were considered per insole to compare NGrF to PFF measurements. The mean error on NGrF over all cycles and insoles was computed and normalized by the maximum NGrF. Bland-Altman plots, RMSE and NRMSE, as well as Pearson correlation coefficient were computed. Results A mean error on NGrF of 31.7 N (SD: 31.9 N) was obtained between insoles and PFF. It corresponds to 3.6% (± 3.7%) of maximum NGrF during stamping. RMSE and NRMSE were of 45.4 N and 5.2% respectively. A 0.99 correlation coefficient was obtained. Bland-Altman plots highlighted that the higher the NGrF the higher the error. Discussion, conclusion Results suggested that the insoles are reliable compared to force platforms. However, highest errors seemed to be associated to calibration issues. Thus, more works must be conducted to evaluate their reliability in long time use, to estimate the intra and interoperator error, and to set up a validated protocol with injured patients. However the biofeedback of NGrF in real time with these embedded sensors open perspectives in rehabilitation.