Multivariate assessment of differences between a neuromuscular electrical stimulation therapy and robotic treadmill training in the rehabilitation of spinal cord injured rats

A study was conducted to evaluate and compare the effects of two different rehabilitation therapies on spinal cord injured (SCI) rats: neuromuscular electrical stimulation which is timed to robotic treadmill training (NMES+RTT) and RTT alone. Several electromyography (EMG) based variables were measured, but most did not change significantly after treatment, contrary to observations of overall qualitative stepping ability. However, when the variables are viewed in multi-dimensional space, there are visible differences between changes after NMES+RTT vs. those after RTT only. Principal component analysis (PCA) and k-means clustering were applied to the multivariate data. The data in principal component space was significantly separated, according to the Euclidean distance. PCA also provided a straightforward tool for selecting which combination of measures to compare. The measures which best separated out the differences between NMES+RTT and RTT were percentage of steps associated with bursts, burst-to-step latency, and the standard deviation of this latency, even though these measures did not always show the greatest statistical significance individually. Thus, the rehabilitative effects of NMES+RTT are not necessarily reflected in individual EMG measures, but rather in a combination of the measures representing a multi-dimensional space.