From Virtual Reality to Ex Vivo Training: An AI-Driven Approach to Enhance Surgical Performance in Neurosurgical Trainees

• Grant number: 25/10401-0
• Support Opportunities: Scholarships abroad - Research
• Start date: November 06, 2025
• End date: August 31, 2026
• Field of knowledge: Health Sciences - Medicine - Surgery
• Principal Investigator: Matheus Fernando Manzolli Ballestero
• Grantee: Matheus Fernando Manzolli Ballestero
• Host Investigator: Rolando Del Maestro
• Host Institution: Centro de Ciências Biológicas e da Saúde (CCBS). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil
• Institution abroad: Neurosurgical Simulation And Artificial Intelligence Learning Centre, Canada

Abstract

Presentation:This project will be carried out at McGill University in Canada, under the supervision of Professor Dr. Rolando Del Maestro, at the Neurosurgical Simulation and Artificial Intelligence Learning Centre (NeuroSim), an international reference in neurosurgical simulation using virtual reality and artificial intelligence. The applicant is an Assistant Professor at the Federal University of São Carlos and a neurosurgeon at the Hospital das Clínicas of the Ribeirão Preto Medical School, University of São Paulo, working in teaching, research, and residency training. The main objective of this proposal is to deepen his theoretical and practical knowledge in advanced surgical simulation and artificial intelligence applied to medical education, aiming to implement innovative technologies in Brazil.Introduction:Neurosurgery is one of the most complex and high-risk medical specialties, requiring highly refined technical training. Traditional models of surgical education have been increasingly replaced by simulation-based and competency-driven training approaches. At NeuroSim, the Intelligent Continuous Expertise Monitoring System (ICEMS) was developed-an intelligent tutoring system that uses deep learning algorithms, such as Long Short-Term Memory (LSTM) networks, to continuously assess technical performance in real time. While ICEMS has already shown effectiveness in virtual reality simulators, it still requires validation in settings with greater tactile and visual realism. For this purpose, an ex vivo subpial resection model using calf brain tissue was developed-a critical technique in epilepsy and brain tumor surgery.Methods:The study will include forty participants across four training levels (medical students, junior residents, senior residents, and specialists), each performing three subpial resections, totaling one hundred and twenty procedures. Instrument movements will be continuously tracked at ten-hertz intervals. Data will be used to train an LSTM-based algorithm to predict expertise levels. Teachability metrics such as acceleration, speed, and jerk will be extracted. Classical statistical analyses (Shapiro-Wilk, Kruskal-Wallis, ANOVA, Mann-Whitney) will be applied, along with machine learning classifiers including K-nearest neighbor, Naive Bayes, discriminant analysis, and support vector machine. Procedure videos will be blindly evaluated by expert raters using the OSATS scale.Expected Results:The ex vivo ICEMS is expected to effectively distinguish skill levels, extract teachable metrics, and provide real-time, expert-based feedback to improve technical performance. The system is also expected to show feasibility for future intraoperative use, enhancing patient safety. Additionally, this project will contribute to the training of undergraduate and graduate students through research opportunities, theses, and dissertations. The acquired knowledge will be transferred to the recently established Ultra-Realistic Neurosurgical Simulation Center at the Ribeirão Preto Medical School, expanding the scientific, educational, and societal impact of this initiative.