Artificial Intelligence-Based Prediction of Surgeon Stress in Robot-Assisted Minimally Invasive Surgery Using ECG Sensor Data

Background/Objectives: Robot-assisted surgery (RAS) has grown rapidly over the past few decades. To determine the effect of high stress levels on the performance of RAS, monitoring some parameters of surgeons is critical. This can be aided by the development of Artificial Intelligence (AI), which has exponentially grown in recent years. This study aims to predict the surgeon’s stress level based on ergonomic, kinematic and physiological parameters of the surgeon obtained in the immediately previous situation during RAS activities. Methods: Physiological data were recorded from surgeons during twenty-six surgical sessions involving twelve participants with different levels of experience and surgical specialties. After dataset generation, two preprocessing procedures (scaling and normalization) were applied to the recorded signals. The processed data were then partitioned into two subsets: 80% of the samples were used for model training and cross-validation, while the remaining 20% were reserved for testing. Six AI approaches were evaluated to build predictive models: multiple linear regression (MLR), a support vector machine (SVM), a multilayer perceptron (MLP), a convolutional neural network (CNN), random forest (RF), and a U-Net algorithm (UNET). These algorithms were trained using the training dataset and subsequently assessed on the independent test set. In addition, after each surgical session, surgeons completed a questionnaire reporting their perceived stress level, which was later compared with the stress estimates generated by the predictive models. Results: The results obtained showed that MLR and scaling pre-processing reached the highest R2 coefficients and the lowest error for each studied parameter. The results of the surgeons’ surveys were highly correlated for microsurgery activities (R2 = 0.7989) and for laparoscopy RAS (R2 = 0.8381). Conclusions: The linear models proposed were correctly validated on cross-validation and the test dataset. This fact demonstrates the possibility of predicting factors that help us to improve the surgeon’s health during RAS.