Rectus femoris EMG clustering, A data-driven management of crouch gait in patients with cerebral palsy (CP)

In 3d-gait analysis in CP, EMG is frequently measured next to kinematics and kinetics but still is little used for surgical treatment indication [1]. Cluster analysis is commonly used to identify trends in datasets [2] and can aid in developing global cerebral palsy Electromyography (EMG) assessment approaches. These approaches, along with kinematics and clinical data, can provide important information for treatment decision-making. How does EMG cluster analysis of Rectus Femoris (Recf) assist in developing better interpretation of gait analysis in patients with CP? Retrospective gait data of patients with CP used for this study was divided in two groups primarily. The group of patients who examined in the gait lab for the first time (E1, before treatment) and those who examined for the second time too (E2, after treatment). After applying the inclusion criteria (having EMG, kinematics, kinetics, physical, and clinical data) 889 and 379 patients were recruited as E1 and E2, respectively. Regularly, EMG data in envelope format [3] from eight lower body muscles were collected, but for the clustering purposes we focused on the Recf. 6 main features (mean, range, max, min and their timing) during whole gait cycle and its subphases were extracted for clustering. Thus, three main clusters (L1, L2 and L3) were identified using PCA, K-means and Elbow algorithms respectively. Additionally, subjects were classified into Crouch, Jump, Recurvatum, Stiff knee and Mild gait according to [4,5]. To further analyze the calculated clusters, we introduced the RG measure as the ratio between the 6 extracted features from Recf and Gastrocnemius lateralis muscles. Kruskal-Wallis test used to compare the RG between E1 and E2 (p-Value = 0.05). The labels were characterized by a peak during loading response (L1), prolonged activity during (terminal) stance (L2), and a significant peak during mid-swing (L3). Notably, L2 contained 76% and 92% of all crouch patients in E1 and E2, respectively (Table 1). Comparing the crouch patients identified as L2-E1 and L2-E2, it was observed that 17 patients had a persistent crouch condition (Group 1), while 8 subjects showed gait improvement in E2 (Group 2). Table 1 shows a significant difference in the minimum and mean of the RG index during terminal stance between these two groups.Download : Download high-res image (169KB)Download : Download full-size image The crouch characteristic of L2 is consistent with the stance-phase prolongation of quadriceps observed in this label [4]. A greater chance of improvement from the crouch might be associated with a higher and lower activity in Gastrocnemius and Rectus during the terminal stance, respectively. EMG clustering of Recf could potentially establish a novel threshold-based treatment decision-making approach for patients with CP having crouch gait.This research was funded by German Research Foundation (DFG) (no: WO 1624/ 8-1).