Estimating total maximum isometric force output of trunk and hip muscles after spinal cord injury

Functional neuromuscular stimulation (FNS) can be used to restore seated trunk function in individuals paralyzed due to spinal cord injury (SCI). Musculoskeletal models allow for the design and tuning of controllers for use with FNS; however, these models often use aggregated estimates for parameters of the musculotendon elements, the most significant of which is maximum isometric force (MIF). Stimulated MIF for individuals with SCI is typically assumed to be approximately 50% of the values exhibited by able-bodied muscles, which itself varies between studies and individuals. A method for estimating subject-specific MIF during dynamic motions in individuals with SCI produced by electrical stimulation has been developed to test this assumption and obtained more accurate estimates for biomechanical analysis and controller design. A simple on-off controller was applied to individuals with SCI seated in the workspace of a motion capture system to record joint angles of three types of trunk motions: forward flexion, left and right lateral bending followed by returning, un-aided, to upright posture via neural stimulation delivered to activate the muscles of the hips and trunk. System identification was used with a musculoskeletal model to find the optimal MIF values that reproduced the experimentally observed motions. Experiments with five volunteers with SCI indicate that an MIF of the 50% able-bodied values commonly used is significantly lower than the identified estimates in 33 of 44 muscle groups tested. This suggests that the strengths of paralyzed muscles when stimulated with FNS have been underestimated in many situations and their true force outputs may be higher than the values suggested for use in simulation studies with musculoskeletal models. These findings indicate that subject-specific musculoskeletal models can more closely mimic the motions of subjects by using individualized estimates of MIF, which may allow the design and tuning of controllers while reducing the time spent with subjects in the loop.