Evaluating different optimization criteria for estimating spine loads and muscle activity when using back-support exoskeletons

Purpose We evaluated how back muscle activity, estimated using different optimization criteria in a commercial musculoskeletal modeling system, corresponds to measured activity. We also examined how the choice of optimization criteria influences estimated intervertebral joint forces (IJFs) during symmetric and asymmetric dynamic lowering/lowering tasks performed with and without two types of back-support exoskeletons (BSEs). Methods We simulated dynamic lowering/lifting tasks in the AnyBody Modeling System™ (AMS) using three available optimization criteria: quadratic, cubic, and min/max, across three intervention conditions, including two BSEs and one control (no BSE) condition. We compared estimated muscle activity with normalized electromyography (nEMG) using maximum normalized cross-correlation and root mean squared error. For each simulation, we estimated peak axial compression and anteroposterior shear IJFs at the lumbar L4/L5 level. Results Quadratic and cubic criteria estimated muscle activity with generally stronger associations and smaller error vs. nEMG. Errors were typically larger during lifting than lowering, and associations were weaker with larger errors in asymmetric compared to symmetric tasks. Estimated IJF magnitudes varied with optimization criteria, with min/max producing the largest forces, followed by cubic and quadratic. Despite these differences in estimated force magnitudes, all criteria indicated a comparable relative reduction in IJFs with BSE use (compared to the control) across all conditions. Conclusion While the selection of an optimization criterion affected the estimated magnitude of spinal loads, it did not change the overall finding of BSE effectiveness. The accuracy of muscle activity estimation using the different criteria was also independent of BSE use.