Clustering analysis of EMG cyclic patterns: A validation study across multiple locomotion pathologies

Among other applications, electromyography (EMG) is used in the assessment of locomotion pathologies to quantitatively document abnormal muscle activation patterns during walking. However, EMG cyclic patterns are affected by high cycle-to-cycle variability. Previous research introduced a clustering approach (CIMAP) to recognize gait cycles with similar EMG onset/offset timings, reducing variability. To demonstrate the feasibility of the approach, the algorithm was validated on healthy subjects. The aim of this study is to extend the validation of the algorithm to multiple locomotion pathologies (both orthopedic and neurological). Gait data of a total of 50 subjects suffering from 5 different locomotion alterations were analyzed, considering 4 lower limb muscles. For each patient, datasets were built grouping EMG cycles with the same number of muscle activations. Then, hierarchical clustering was applied to each dataset and cycle-to-cycle variability was calculated for each cluster. Our results showed that CIMAP reduced the median variability below 5% of the gait cycle, for all the considered pathologies. Analyzing the number of clusters obtained we found that, in the great majority of cases, gait cycles cannot be bunched into a single group, but rather 2 or more clusters are necessary. As a consequence, the cluster representative elements, calculated by averaging cycles belonging to the same cluster, provide more trustworthy information for the clinician than indiscriminately averaging all cycles from a dataset.