Fig. 6: Space-by-time decomposition.
a Result of the space-by-time decomposition of EMG data of neonates using two modules. Left column: muscle synergies (w1 and w2, from top to bottom), each column represents a single neonate, and each row represents a muscle (BF: Biceps Femoris, RF: Rectus Femoris, LG: Lateral Gastrocnemius, TA: Tibialis Anterior, CL: Contralateral Leg), the intensity of the color is proportional to the muscle weight, the average (+SD) values are illustrated through gray bars on the right. Upper row: basic activation patterns (c1 and c2, from left to right), each colored line represents a single neonate, black lines indicate the average across subjects. Middle panels: elements of the activation coefficients matrix (a11, a12, …, a22), each bar represents a single stride (s1-s7) of each neonate (in different colors). The modules of each subject are grouped in order to attain the minimum distance (1-scalar product) between the basic activation patterns of all subjects. Since for the space-by-time decomposition there is not a unique relationship between synergies and basic patterns, which are related by the NxN matrices of activations coefficients, after sorting the basic activation patterns, we sorted also the muscle synergies in order to obtain the resulting activation coefficient matrix as close as possible to a diagonal matrix. b Space-by-time decomposition of EMG data of neonates using four modules, same format as panel a. c Space-by-time decomposition of EMG data of adults using four modules, same format as panels a and b. Note that, despite four modules in neonates account on average for 69% (CI95% [66%, 71%]) of the variance of the EMG data [in neonates two modules account on average for 41% (CI95% [39%, 44%]) of the variance, and in adults four modules account on average for 78% (CI95% [75%, 81%]) of the variance], these modules are visibly less consistent across strides and across subjects (the average synergies across subjects are flatter and the non-diagonal elements of the activation coefficients matrices are noticeably higher) compared with the decomposition with two modules in neonates (a) and four modules in adults (panel c). N6 data are missing because this neonate performed 5 consecutive strides, but did not have the seven strides necessary for factor analysis (see Methods).
