Health Canada (Commonwealth Union) – Researchers of a study point out that Virtual reality simulators offer a valuable means for learners to enhance their technical skills rapidly and without endangering patients. In the context of neurosurgery, they enable medical students to practice intricate procedures before performing them on real patients. When integrated with artificial intelligence, these instructional systems can provide personalized feedback akin to that of a human instructor. They can pinpoint areas where students require improvement and offer suggestions to help them achieve expertise.
Nonetheless, a recent study conducted by the Neurosurgical Simulation and Artificial Intelligence Learning Centre at The Neuro (Montreal Neurological Institute-Hospital) of McGill University underscores the continued necessity of human instruction for detecting and mitigating unintended, occasionally adverse, alterations in neurosurgeon behavior following AI training in virtual reality.
In this study, 46 medical students engaged in a virtual reality-based tumor removal procedure. Half of them were randomly assigned to receive guidance from an AI-driven intelligent tutor known as the Virtual Operative Assistant (VOA), which employs a machine learning algorithm to impart surgical techniques and deliver personalized feedback. The remaining half constituted a control group that received no feedback. Subsequently, the students’ performance was compared against benchmarks established by a panel of experienced neurosurgeons.
Comparing the outcomes, it was observed that AI-instructed students caused 55 percent less harm to healthy tissues compared to the control group. They also displayed a 59 percent reduction in the average distance between instruments held in each hand, along with a 46 percent decrease in the maximum force applied—both of which are critical safety indicators.
However, AI-instructed students also exhibited certain adverse results. Notably, their dominant hand movements displayed a 50 percent lower velocity and a 45 percent lower acceleration than the control group, resulting in less efficient procedures. The speed at which they removed tumor tissue was likewise 29 percent slower in the AI-instructed group compared to the control group.
These unintended outcomes underscore the significance of human instructors in the learning process, as they play a vital role in fostering both safety and efficiency among students.
The imperfect nature of AI systems was pointed out by Ali Fazlollahi, a medical student researcher serving as the study’s primary author at the Neurosurgical Simulation and Artificial Intelligence Learning Centre. He further indicated that attaining expertise will still necessitate a certain degree of guidance from a seasoned practitioner. Implementing AI programs will empower learners to assess their proficiency and allocate their in-operation learning time more efficiently with instructors, focusing on their personalized learning objectives. They are presently in the process of developing an optimal hybrid instructional approach in an ongoing crossover trial.
Fazlollahi underscores that his findings extend beyond the realm of neurosurgery, as many of the same principles find application in other domains of skill development.
“This includes surgical education, not just neurosurgery, and also a range of other fields from aviation to military training and construction,” he added. “Using AI alone to design and run a technical skills curriculum can lead to unintended outcomes that will require oversight from human experts to ensure excellence in training and patient care.”
“Intelligent tutors powered by AI are becoming a valuable tool in the evaluation and training of the next generation of neurosurgeons,” said Dr. Rolando Del Maestro, who is the senior author of the study. “However, it is essential that surgical educators are an integral part of the development, application, and monitoring of these AI systems to maximize their ability to increase the mastery of neurosurgical skills and improve patient outcomes.”
Published on September 19, 2023, in the Journal of the American Medical Association (JAMA Network Open), this research was financially supported by the Franco Di Giovanni Foundation, the Brain Tumour Foundation of Canada Brain Tumour Research Grant, a Medical Education Research Grant from the Royal College of Physicians and Surgeons of Canada, as well as the Montreal Neurological Institute-Hospital.
