Purpose: Over 90,000 osteoarthritis (OA) related TKR surgeries take place across the UK annually, with patients undergoing regular post-surgery physiotherapy that is reliant on home-based exercise rehabilitation and driven by personalised self-management. With poor patient adherence that is difficult to ascertain, clinicians who are challenged to optimise patient outcomes are unable to determine whether improvements (or lack of) can be attributed to an exercise intervention or (non) adherence. There is a clear need for enhanced forms of objectively monitoring patient adherence to home based exercise rehabilitation, providing valuable biomechanical knowledge to clinicians to guide personalised exercise prescription. This could provide rigorous adherence measurements, optimise the rehabilitation process, reduce NHS burden and improve patient satisfaction. This research aims to determine whether the performance of 4 rehabilitation exercises, routinely prescribed to OA patients following TKR, can be objectively distinguished using inertial measurement sensors (IMU's) placed on the lower limbs. Methods: 5 healthy participants (4 males, 1 female; mean age 32.6 ± 11.1 years, height 1.79 ± 0.14 m and mass 82.88 ± 15.93 kg) performed a battery of early phase knee rehabilitation exercises based on the Taxonomy for RehAbilitation of Knee conditions (TRAK). Data was collected for multiple exercises with participants wearing a range of time synchronised biomechanical measurement systems. This study focused on the performance of 1) Knee Flexion in sitting 2) Knee Extension 3) Single Step Down and 4) Sit to Stand, with each participant performing 4 repetitions per exercise and the data collected using lower body IMU sensors. These were placed on the pelvis and bilateral thigh, shank and feet (Xsens, Holland; sampling at 60 Hz). Anthropometric measurements for each participant were combined with IMU data during a static calibration to define the biomechanical model (MVN Studio). 3D hip, knee and ankle joint angles were calculated using the Euler sequence ZXY using the ISB based coordinate system. Joint angle data were processed in Python, with exercise repetitions defined using a detect peaks algorithm. 3D angle data were formatted in Excel, time normalised to 101 points and then Principal Component Analysis (PCA) was performed (Matlab, Mathworks), reducing all joint angle waveforms into new uncorrelated principal components via an orthogonal transformation. Scatterplots of PC1 versus PC2 were used to visually inspect for clustering between the PC values for the 4 exercise groups. A one way ANOVA (SPSS, IBM) was performed on the first 3 PC values (ranked by percentage variance accounted for) for the 9 variables under analysis, with an a priori alpha level of significance set at 0.05. Games-Howell post hoc tests identified variables that were significantly different between exercises. Results: The PC scatterplot representing the hip flexion-extension waveforms produced the most prominent clustering, with all 4 exercise groups easily distinguishable (Fig.1). Whilst multiple statistically significant differences were found between pairs of exercises for individual PC values, only one PC value was statistically different across all exercise pairings (PC1, knee flexion-extension waveform). Conclusions: This study demonstrates the potential to objectively distinguish between different knee rehabilitation exercises using IMU sensors and PCA. It would appear that flexion-extension angles at the hip and knee are most suited for accurate exercise classification and require further investigation. Future work will focus on increasing the healthy cohort sample size and generating a post TKR patient cohort to identify whether similar differentiation between exercises can be established in a pathological cohort, and whether there are functional difference between healthy and post-TKR patients that could be used to map patient progress.
Search from 250 M+ research papersSearch from 250 M+ research papers
Open Access
Mar 1, 2019
Jiahui Li + 14
