MULTIMODAL STATISTICAL MODELING AND APPLICATION OF ROBOT-ASSISTED SURGERY PERFORMANCE EVALUATION BASED ON STACKED REGRESSION

Abstract

With the wide application of Robot-Assisted Surgery (RAS) technology, accurate and objective surgical performance evaluation has become the key to optimizing surgical effects and reducing medical risks. Aiming at the problems of strong subjectivity and insufficient accuracy of existing evaluation methods, this paper integrates electroencephalography (EEG) and eye-tracking multimodal data to construct a set of robot-assisted surgery performance evaluation models. First, the original data are normalized, feature extracted and dimensionality reduced to eliminate data heterogeneity and redundancy; then, Support Vector Regression (SVR) and Multilayer Perceptron (MLP) are used as base learners, combined with Newton's iteration method to optimize parameters, to construct a stacked regression model, and linear regression and LSTM models are introduced as controls to carry out comparative experiments; finally, the model performance is quantitatively evaluated by indicators such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and F1-Score. The experimental results show that the proposed stacked regression model is significantly superior to the control models in evaluation accuracy, with RMSE 0.23 lower than that of traditional linear regression and MAE 0.11 lower, which can accurately describe the cognitive and behavioral states of doctors during surgical operations. This study provides a reliable technical method and data support for robot-assisted surgery performance evaluation, and has important theoretical value and practical significance for promoting the optimization of surgical technology and improving medical quality.