Figure 7

Statistical analysis of EMG patterns during SW and FW using non-negative matrix factorization. (A) Basic activation patterns in TD children and in children with HE and DI. Each curve represents the mean (across strides) pattern for an individual child, average patterns across children are illustrated with black lines. Data for the trailing (dark blue lines) and leading (light blue lines) limb are shown separately for SW, data for the right leg are shown for FW (green lines). The basic patterns from the same cluster (c1–c4; based on k-means clustering on muscle synergies) are plotted in a “chronological” order (with respect to the timing of the main peak in FW). (B) Average (± 95%CI, z-transformed, across subjects) similarity (cosα) of the basic activation patterns (c1–c4, from top to bottom) in FW with the basic patterns in trailing and leading limbs during SW. Red lines denote significant (post-hoc Tukey–Kramer multiple comparison p < 0.05) differences between TD and HE or DI children. (C) Muscle synergies (weighting coefficients w1–w4) of corresponding basic patterns (c1–c4) plotted in colour scale. Each row represents a condition (leading/trailing limb for SW and right limb for FW) and each column represents a muscle, the intensity of the colour is proportional to the muscle weight. Red lines denote significant (post-hoc Tukey–Kramer multiple comparison p < 0.05) differences between conditions. (D) Polar plots of the centre of activity (CoA) of “adaptable” c2 and c4 activation patterns for each group. Polar direction denotes the relative time over the gait cycle (time progresses counter-clockwise), radius of the vector denotes the average (across subjects) maximum amplitude of the basic pattern and the width of the sector denotes angular SD. Red lines denote significant (circular Watson-Williams test with Bonferroni correction for multiple comparisons p < 0.05) differences across conditions.