Prediction of localized muscle fatigue in minimally invasive urological surgery

Introduction This study aims to predict localized muscle fatigue during minimally invasive urological procedures, including both laparoscopic and robotic-assisted techniques. Material and Methods For conventional laparoscopy, the Olympus Visera III (Olympus) and Storz instruments were used, the Versius robotic platform (CMR Surgical) was used to RAS. A single surgeon performed two total nephrectomies on a porcine model employing both surgical techniques. Localized muscle fatigue was assessed bilaterally in the following muscle groups: brachioradialis, erector spinae, medial gastrocnemius, middle trapezius, upper trapezius, triceps brachii and vastus lateralis. Subsequently, the following artificial intelligence predictive techniques were applied: Multilayer Perceptron (MLP), Random Forest (RF), Convolutional Neural Networks (CNN), Multiple Linear Regression (MLR), Partial Least Squares Regression (PLS), Support Vector Machine (SVM) and the U-Net (UNET). Results For both surgical techniques, the most effective performance was achieved by RF, MLR and PLS (R2>0.75), followed by SVM and UNET (R2>0.45). The least favorable results were reached for CNN and MLP (R2≈0). Generally, superior outcomes were achieved with RAS in almost all evaluated muscles. The highest predictive accuracy was found for the left medial gastrocnemius and the right upper trapezius and vastus lateralis (R2>0.94). Conclusions The obtained results demonstrate the robustness and accuracy of linear predictive models. This finding could serve as a foundation for further investigation into surgeons’ ergonomic and physiological risks and its prevention.