Two weeks ago, a group of researchers from the Neurosurgical Simulation and Artificial Intelligence (AI) Learning Centre, Department of Neurosurgery, Montreal Neurological Institute and Hospital of McGill University, Canada, published a paper on the application of machine learning (ML) techniques in analyzing performances on virtual reality (VR) simulators.
Fundamentally, VR simulators account for all the movements and forces acted upon them, this generates a large amount of data in the process. Thus, there is a growing trend to use ML algorithms to evaluate some of these datasets, in order to better understand, assess and train psychomotor performances. The research team from McGill had noted the potential of such development, in expanding the number and complexity of studies that could possibly span across multiple disciplines including medicine, computer science and education in the near future.
This brought them to develop a checklist, one which looks into surgical training, called the Machine Learning to Assess Surgical Expertise (MLASE). The research team hopes that it will provide some form of guidelines to ascertain quality when generating or reviewing VR manuscripts which also incorporate the use of ML in assessing surgical expertise.
Taking into consideration of the differences that each discipline has in result reporting, communication and knowledge transfer, this checklist comes with sub-components and a total score which enable future researchers to assess the overall quality or zoom into a particular part.
Joint forces between VR and AI in medicine
Dr. David M. Axelrod, Clinical Associate Professor of Pediatrics Cardiology at Stanford School of Medicine candidly compared the intersection of VR and AI to an analogy between DJs, bank robbers, preachers and mom taking off your sweater at the recent AIMed Cardiology conference. Dr. Axelrod described VR as the melting of arts and science. Previously, he had to draw and explain the surgery he was about to perform on patients. With VR, the drawings are now 3D and colored which patients can interact with.
Indeed, AIMed had extensively explored the impact of VR on medicine earlier. From the feasibility of using VR to treat addiction, if it can be a safe form of training for surgeons, to its practicality to be widely implemented in hospitals to benefit a larger group of patients. The joint forces between two new forms of technologies, however, is creating a whole new landscape.
For example, Dr. Axelrod is collaborating with Dr. Alison Marsden, the Associate Professor of Pediatric Cardiology, Bioengineering and Mechanical Engineering at Stanford University to develop models of blood flow and physiology for real-time flow visualization and feedback during medical training. Presently, they are working with computer scientists to sift through data from medical imaging to create a system of new models that is able to view changes of blood flow via VR.
Dr. Axelrod also plans to create models of heart with defects and fit them into the existing blood flow models for surgical planning purposes. He believes it will provide surgeons with insights on how the surgery may turn out to be when the plan is taken to action. Besides, Dr. Axelrod is confident in the opportunities that VR and AI can offer to exert a change in the clinical, education and research aspects of medicine.
He cited in a recent randomized, controlled and crossover study of pediatric residents. One group of the students took interventional VR sessions while the other group attended a lecture-based session. Knowledge tests were performed on both groups and subsequently at one, three, and six months intervals. It showed that students who had attended the VR session had better performance and higher retention rate of the knowledge that they had learned. “Hence, VR is not just a cool toy, but a cool toy that is really making a difference,” Dr. Axelrod said.
A science writer with data background and an interest in the current affair, culture,