Machine Learning-Based Classification and Prediction of Viscoelastic Creep in Lumbar Tissues

Sustained trunk flexion can cause lumbar spinal creep, a risk factor for low back disorders, but measuring creep directly is impractical outside the laboratory. This study explored whether machine learning (ML) can classify and predict creep from readily available individual and exposure characteristics. We pooled data from three published studies (44 participants, seven flexion and work-rest conditions). Creep was quantified as the change in peak lumbar flexion angle and EMG-off angle. Random forest, support vector machine, and multilayer perceptron classifiers separated low from high creep with accuracies of 0.83 to 0.85 and recall of 0.91 to 0.94. For creep magnitude, Gaussian process regression achieved the best interval coverage (PICP 91.76%), with an RMSE of 1.79°. These results suggest ML can support early monitoring of creep risk using accessible data.