Deactivation and collective phasic muscular tuning for pointing direction: Insights from machine learning

Comparing arm to whole-body motor control disambiguates age-related deterioration from compensation

“…Our rationale was the following: If the algorithm can successfully separate the data of younger and older adults, using antigravity muscle activation patterns, this would demonstrate that important information is contained in those muscles regarding age-related modifications of movement control. For more details on similar use and operation of machine learning algorithms on EMG signals, please see (Chambellant et al, 2024;Thomas et al, 2023;Tolambiya et al, 2011). Here we present the results of a Linear Discriminant Analysis (LDA, Johnson & Wichern, 1988) but we verified that we obtained similar conclusions with two other algorithms, namely the Quadratic Discriminant Analysis (QDA, Cover, 1965) and the Support Vector Machine (SVM, Vapnik and Lerner, 1965).…”

“…This analysis indeed revealed that antigravity muscles contained important information, allowing separating age-groups with some of the best success-rates (see Figure 6 for results regarding LDA accuracy). The vastus lateralis (VL) and the spinal erectors on L1 (ESL1) achieve classification accuracies of 57.72% and 59.51% respectively (considering that these classifications are significantly better than chance if they are above 52.5% according to a fairness test).Thus, building on the complementary results of the theory-driven approach (Chambellant et al, 2024;Gaveau et al, 2021;Poirier et al, 2022Poirier et al, , 2024Thomas et al, 2023) and the present data-driven control analysis (Figure 6), in the following we focus the analysis on…”

“…It was recently shown that the phasic EMG activity of antigravity muscles, those that pull against the gravity vector, consistently exhibits negative epochs (Chambellant et al, 2024;Gaveau et al, 2021;Poirier et al, 2022;Thomas et al, 2023) when the arm acceleration sign is coherent with the gravity acceleration sign (i.e., in the acceleration phase of downward movement and in the deceleration phase of upward movements). This observation likely reflects an optimal predictive motor strategy where muscle activity is decreased when gravity assists arm movements, thereby discounting muscle effort (Gaveau et al, 2021).…”