Quantitative Estimation of Movement Progress during Rehabilitation after Knee/Hip Replacement Surgery
MetadataShow full item record
Mobility improvement for patients is one of the primary concerns of physiotherapy rehabilitation. In a typical physiotherapy session, the patient is instructed to perform multiple exercises, based on a specific regimen recommended by the physiotherapist for each patient. The physiotherapist then evaluates the patient's progress based on his or her performance during the exercises. Providing the physiotherapist and the patient with a quantified and objective measure of progress, based on both individual exercises and the exercise set, can be beneficial for monitoring the patient's performance. The quantified measure can also be beneficial when the physiotherapist is not available, e.g., crowded gym or rehabilitation at home. In this thesis, two approaches are introduced for quantifying patient performance. One approach describes the movement timeseries by statistical measures and the other by a stochastic model. Both approaches formulate a distance between patient data and the healthy population as the measure of performance. Distance measures are defined to capture the performance of one repetition of an exercise or multiple repetitions of the same exercise. To capture patient progress across multiple exercises, a quality measure and overall score are formulated based on the distance measures and are used to quantify the overall performance for each session. The proposed approaches are compared to several existing approaches, including sample distribution approaches (two sample kernel), classifier-based approaches (Naive Bayes, Support Vector Machines, and Kullback-Leibler Divergence), and dynamical movement primitives. In their original formulation, existing approaches are not capable of estimating measures of performance for multiple exercises. Therefore, the measures of performance for multiple repetitions of the same exercise are estimated using the existing approaches, while the formulation proposed in this thesis is used to estimate the overall performance for multiple exercises in one session. The effects of different variabilities in human motion on the performance of the proposed approaches and the comparison approaches are investigated with both synthetic and patient data. The patient data consists of rehabilitation data recorded from patients recovering from knee or hip replacement surgery, the associated exercise regimen and physiotherapist evaluations of progress. The methods are evaluated quantitatively based on correlation between methods, correlation with exercise regimen difficulty, and qualitatively based on the patients' medical charts. The proposed approaches are capable of capturing the trend of progress for the synthetic dataset and are superior to the existing approaches in presence of multiple sources of variability. For patient data, the proposed approaches correlate moderately with the score obtained from the exercise regimen, and qualitatively correspond with the patients' medical charts. The results indicate that the quantified measures of progress obtained from the proposed approaches are promising tools for supporting physiotherapy practice through monitoring patient progress.