Muscle-level evaluation of the minimum muscle-stress-change model in human three-joint reaching using anatomically...

This study examined whether muscle-level optimization can account for both arm movements and muscle recruitment in planar three-joint reaching. To extend previous movement-level analyses based on simplified musculoskeletal representations, anatomically detailed three-joint arm models comprising 19-26 muscles with nonlinear, joint-angle-dependent moment arms were used. This extension enabled direct evaluation of whether muscle-level computational models can resolve highly redundant muscle-tension distributions. The reaching task was designed with eight directions at ${\text{45}}^{\circ }$ intervals in the horizontal plane, allowing evaluation with reduced directional bias. Several computational models were compared, with particular emphasis on the minimum muscle-stress-change (MSC) framework. The quadratic ${\text{MSC}}_2$ , cubic ${\text{MSC}}_3$ , and minimum muscle-tension-change (MTC) models reproduced the measured arm movements reasonably well; among the muscle-level models, ${\text{MSC}}_2$ provided the most accurate predictions, particularly for arm posture and the direction-dependent wrist-joint contribution. This level of accuracy justified evaluation of the predicted muscle-recruitment patterns. The MSC framework preferentially recruited muscles with larger physiological cross-sectional areas and avoided excessive loading of thinner muscles, whereas the MTC model was more strongly influenced by moment-arm values and sometimes assigned relatively large tensions to thinner muscles. The ${\text{MSC}}_3$ model recruited nearly all muscles in the 26-muscle expanded arm model but showed slightly lower movement-reproduction accuracy than the ${\text{MSC}}_2$ model. Thus, ${\text{MSC}}_2$ provided a balanced account of task-space behavior, joint-space coordination, and muscle recruitment. These findings extend muscle-stress-based optimization from movement reproduction to physiologically interpretable recruitment prediction and suggest that synergy-like cooperative patterns can emerge from anatomical constraints and stress-based optimization.