Predicting Surgical Experience After Robotic Nerve-sparing Radical Prostatectomy Simulation Using a Machine...

Area of Science:

• Urology
• Surgical Simulation
• Machine Learning

Background:

• Objective evaluation of surgical performance is crucial for training and quality assessment.
• Machine learning offers objective tools for assessing operative performance in urological procedures.
• Predicting surgeon caseload and expertise is vital for optimizing surgical training and patient outcomes.

Purpose of the Study:

• To develop machine learning (ML) methods for predicting surgeon caseload in nerve-sparing robot-assisted radical prostatectomy (RARP).
• To identify key metrics indicative of surgical expertise using a validated hydrogel-based simulation platform.
• To establish a predictive algorithm for surgical experience in RARP.

Main Methods:

• Collected video, robotic kinematics, and force sensor data from 35 urologists during simulated RARP.
• Annotated surgical gestures and derived objective performance indicators from kinematic data.
• Utilized logistic regression, support vector machine, and k-nearest neighbors models, optimizing with recursive feature elimination.

Main Results:

• Logistic regression with recursive feature elimination achieved the highest Area Under the Curve (AUC) of 96% in predicting surgical experience.
• Combined data including objective performance indicators, gestures, and force metrics yielded high prediction accuracy (up to 94%).
• The study identified key contributory features across different ML models for predicting caseload.

Conclusions:

• A novel ML-based algorithm combining objective performance indicators, gesture analysis, and force metrics can predict surgical experience in RARP with 96% AUC.
• The developed algorithm effectively discriminates between surgeons with low and high caseloads in a standardized simulation environment.
• This approach provides an objective measure of surgical expertise for nerve-sparing robot-assisted radical prostatectomy.