Applying artificial intelligence on EDA sensor data to predict stress on minimally invasive robotic-assisted surgery

Area of Science:

• Robotics and Surgical Technology
• Human Factors and Ergonomics
• Biomedical Engineering

Background:

• Minimally invasive robotic surgery offers precision but can induce surgeon stress.
• Monitoring surgeon stress is crucial for patient safety and well-being.
• Objective stress assessment in surgical environments remains a challenge.

Purpose of the Study:

• To predict surgeon stress levels during robotic surgery.
• To correlate stress predictions with ergonomic and physiological parameters.
• To validate predictive models against surgeon self-reported stress.

Main Methods:

• Collected kinematic and physiological data (electrodermal activity, blood pressure, body temperature) from 11 surgeons across 26 robotic surgery sessions.
• Applied data preprocessing (scaling, normalization) and machine learning models (MLR, SVM, MLP).
• Validated models using 80% training/cross-validation and 20% test data splits, comparing predictions with post-session surgeon surveys.

Main Results:

• Multiple linear regression (MLR) with scaled preprocessing yielded the best predictive performance (highest R² and lowest error).
• Predictive models showed a high correlation with surgeon-reported stress levels (R² = 0.8253).
• Validated linear models demonstrated predictive accuracy on unseen data.

Conclusions:

• The study successfully developed and validated linear models for predicting surgeon stress in robotic surgery.
• Objective prediction of surgeon stress is feasible using ergonomic and physiological data.
• These findings can inform strategies to enhance surgeon health and improve surgical outcomes.