Classification of gait muscle activation patterns according to knee injury history using a support vector machine approach

Abstract

Abnormal muscle activation patterns during gait following knee injury that persist past the acute injury and rehabilitation phase (>three years) are not well characterized but may be related to post-traumatic knee osteoarthritis. The aim was to characterize the abnormal muscle activity from electromyograms of five leg muscles that were recorded during treadmill walking for young adults with and without a previous knee injury 3–12 years prior. The wavelet transformed and amplitude normalized electromyograms yielded intensity patterns that reflect the muscle activity of these muscles resolved in time and frequency. Patterns belonging to the affected or unaffected leg in previously injured participants and patterns belonging to a previously injured vs. uninjured participant were grouped and then classified using a principal component analysis followed by a support vector machine. A leave-one-out cross-validation was used to test the model significance and generalization. The results showed that trained classifiers could successfully recognize whether muscle activation patterns belonged to the affected or unaffected leg of previously injured individuals. Classification rates of 83% were obtained for all subjects, 100% for females only, indicating sex-specific knee injury effects. In contrast, it was not possible to discriminate between patterns belonging to the previously injured legs or dominant legs of control subjects. For females, the injured leg showed a stronger muscle activity for hamstring muscles and a lower activity for the vastus lateralis. In conclusion, systematic knee injury effects on the neuromuscular control of the knee during gait were present 3–12 years later.