A new study from the FAMU-FSU College of Engineering, published in JAMA Surgery, has the potential to improve surgical training. Led by Suvranu De, the college’s Google-endowed dean, the team developed an AI-powered application that analyzes surgical technique videos and provides feedback, enhancing the training process for surgeons.
Image Credit: Gorodenkoff/Shutterstock.com
If you have ever had to make the challenging decision to undergo surgery, you probably had many questions, including concerns about the skill and experience of the surgeon performing the procedure. To this end, researchers from the FAMU-FSU College of Engineering have found a way to potentially improve surgical training.
The more training and feedback surgeons-in-training receive, the more their skills will improve. We have established a cutting-edge video-based assessment network (VBA-Net) that is a major step in the direction of automating the evaluation of surgical skills effectively. This system uses state-of-the-art deep learning models for formative and summative evaluations that foster skill development.
Suvranu De, Dean, FAMU-FSU College of Engineering, Florida State University
VBA-Net is an AI model designed to differentiate between expert and novice surgeons by analyzing full-length videos of real surgical procedures. It offers learners final scores and real-time feedback, automating the assessment process traditionally conducted by trained proctors.
This platform integrates deep neural network (DNN) technology with video-based surgical assessment to provide immediate feedback to aspiring surgeons. DNNs, a type of artificial intelligence that simulates the complexity of the human brain, help personalize the learning experience, making feedback more relevant and tailored to individual needs.
This tool can offer valuable support to evaluators and has the potential to ensure greater consistency in assessments. Our objective is to streamline the evaluation process by guiding trainees in their focus to the most critical facets of a surgical procedure.
Suvranu De, Dean, FAMU-FSU College of Engineering, Florida State University
The DNN technology used in VBA-Net includes Explainable Artificial Intelligence (XAI), which makes the inner workings of the AI model more transparent and understandable for users. This feature enhances trust in the results produced by machine learning algorithms and requires only minimal hardware and a standard camera setup.
De's research supports the American Board of Surgery’s initiative to integrate video-based assessment (VBA) into surgical training. The Board began a pilot program in 2021 to standardize VBA, and De's innovative approach aligns AI directly with this assessment method.
“We hope the insights from this research can pave the way for integrating this technology in training and credentialing programs in the next five to ten years. Our ultimate aspiration is to enhance patient outcomes, save lives, and cultivate more well-trained surgeons in the future,” De stated.
The study was conducted in collaboration with Erim Yanik, a postdoctoral researcher at the FAMU-FSU College of Engineering, and Dr. Steven Schwaitzberg, chair of surgery at the University of Buffalo’s Jacob’s School of Medicine and Biomedical Sciences.
Journal Reference:
Yanik, E., et. al. (2024) Deep Learning for Video-Based Assessment in Surgery. JAMA Surgery. doi:10.1001/jamasurg.2024.1510