This study aimed to create models of knee instability using inertial sensors. Seventy older adults, aged 71.2 ± 6.6 years, participated in this study. The acceleration and angular velocity were obtained using inertial sensors placed on the lower limb and five physical therapists assessed the knee instability of the subjects hierarchically. Thirty-nine variables, including the amplitudes of the acceleration, and the angular velocity for each sensor and for the three axes, were extracted. Multivariate ordered logistic regression analysis was performed to identify the factors affecting knee instability in the observational assessment. And the five-graded knee instability assessment models were created using variables that were excluded if they correlated with each other. The results indicated that the increased lateral acceleration [odds ratio (OR) = 3.51, p<0.01] and the decreased vertical acceleration [OR = 0.43, p<0.01] of the distal femur in the stance phase, and the increased angular velocity of the proximal tibia in the stance phase [OR = 1.91, p<0.01] were significant predictive factors for knee instability. This study found that all of the above three indicators were selected in all models driven by stepwise selection in each leave-one-subject-out cross-validation. Therefore, these factors could be used for knee instability evaluation model.
View full abstract