Lower Knee Joint Loading by Real-time Biofeedback Stair Walking Rehabilitation for Patients With MCKOA

Overview

This study will establish a machine-learning algorithm to predict KAM using IMU sensors during stair ascent and descent; and then conduct a three-arm randomized controlled trial to compare the biomechanical and clinical difference between patients receiving a course of conventional laboratory-based stair retraining, sensor-based stair retraining, and walking exercise control (i.e., walking exercise without gait retraining). The investigators hypothesise that the wearable IMUs will accurately predict KAM during stair negotiation using machine-learning algorithm, with at least 80% measurement agreement with conventional calculation of KAM. The investigators also hypothesise that patients randomized to the laboratory-based and sensor-based stair retraining conditions would evidence similar (i.e., weak and non-significant differences) reduction in KAM (primary outcome) and an improvement of symptoms (secondary outcomes), but that these subjects would evidence larger reductions in KAM than subjects assigned to the walking exercise control condition.

Conventionally, gait retraining is necessarily implemented in a laboratory environment because evaluation of biomechanical markers, such as KAM, requires sophisticated motion capturing system and force plates. With advancement of wearable sensor technology, it is possible to measure gait biomechanics and provide real time biofeedback for gait retraining using inertial measurement unit (IMU), which is a lightweight and portable wireless device. In an ongoing government funded project, the investigators have developed IMU embedded footwear that measures KAM during level ground walking. The investigators have compared Least Absolute Shrinkage and Selection Operator (LASSO) regression and Random Forest in the prediction of KAM from IMU recordings. The investigators found that Random Forest could provide much higher KAM prediction accuracy than LASSO regression. The agreement between conventional laboratory-based and sensor-based measurement of KAM was approximately 90%. Based on investigators' previous research work, it is meaningful to extend the newly developed technology for KAM measurement during stair ascent and descent without the use of laboratory equipment. With the wearable sensors connected to the smartphones, gait retraining outside laboratory environment will become feasible but the effects of gait retraining using wearable sensors have not been directly verified. Given these considerations, this project has two primary aims. The investigators will: (1) first establish a machine-learning algorithm to predict KAM using IMU sensors during stair ascent and descent; and then (2) conduct a three-arm randomized controlled trial to compare the biomechanical and clinical difference between patients receiving a course of conventional laboratory-based stair retraining, sensor-based stair retraining, and walking exercise control (i.e., walking exercise without gait retraining). Primary hypothesis Hypothesis 1: The wearable IMUs will accurately predict KAM during stair negotiation using machine-learning algorithm, with at least 80% measurement agreement with conventional calculation of KAM. Hypothesis 2: Patients randomized to the laboratory-based and sensor-based stair retraining conditions would evidence similar (i.e., weak and non-significant differences) reduction in KAM (primary outcome) and an improvement of symptoms (secondary outcomes), but that these subjects would evidence larger reductions in KAM than subjects assigned to the walking exercise control condition.