Skills evaluation, tailored feedback: McGill AI project could change the way brain surgeons are trained

Alexander Winkler-Schwartz, a neurosurgery resident and PhD candidate, poses in the lab with a NeuroVR neurosurgical simulator at McGill University, on July 31, 2019.

CHRISTINNE MUSCHI/THE GLOBE AND MAIL

Alexander Winkler-Schwartz focuses on the computer-generated brain on the screen while, below, his hands gently remove the virtual brain tumour inside the mannequin’s head.

An artificial intelligence algorithm tracks the neurosurgery resident’s every movement – ready to classify his performance as part of a research project at McGill University, where intelligent machines are learning to rank people based on how deftly they take away the tumour.

It’s part of a wider effort to harness the power of technology to improve medicine. Artificial intelligence is already helping monitor the vital signs of babies in intensive-care, and robots are a fixture in operating rooms.

The algorithm in Montreal is correct in determining whether the virtual surgeon is an expert, a beginner, or somewhere in between 90 per cent of the time, based on a study of 250 simulated operations by 50 different people, according to a study published in the open-source Journal of the American Medical Association on Friday that Dr. Winkler-Schwartz was a lead author on.

An artificial intelligence algorithm tracks Dr. Winkler-Schwartz’s every movement – ready to classify his performance as part of a research project.

CHRISTINNE MUSCHI/THE GLOBE AND MAIL

The technology could have important applications for training and testing aspiring neurosurgeons, since it can evaluate their skills and give them tailored feedback on how to more closely match the movements of an expert.

“If you give [the machines] enough information about how the expert’s hands move, they can determine what makes an expert an expert,” said Rolando Del Maestro, a neurosurgeon with more than 30 years of experience who was involved in the study.

Tiny motors inside the surgical instrument handles create resistance so users can feel the thickness of the virtual tumour compared to the brain tissue as they try to remove it. Expert surgeons use more gentle pressure and have more dexterous left hands, Dr. Del Maestro said.

Montreal neurosurgery students and researchers have used the simulator for the last decade or so, although it’s not yet integrated into the broader curriculum.

Tiny motors inside the surgical instrument handles create resistance so users can feel the thickness of the virtual tumour compared to the brain tissue as they try to remove it.

CHRISTINNE MUSCHI/THE GLOBE AND MAIL

The latest research using machine learning is significant because it allows a user’s level of skill to be classified in a relatively objective fashion, according to Abhaya Kulkarni, a neurosurgeon and senior scientist at Toronto’s Hospital for Sick Children who was not involved in the study.

Until now, neurosurgery has been an apprenticeship model. Students rely on verbal and written feedback from their mentors, who look for cues like how comfortable students seem with the instruments and how efficiently they operate, Dr. Kulkarni said.

“Right now, for the most part, it does require us observing and making a subjective assessment,” he said.

In the future, once it’s established that prowess in the simulator is tied to success in the operating room, these algorithms could test surgical residents before they become certified, Dr. Del Maestro suggested.

Dr. Kulkarni agreed that’s a possibility, adding right now there’s no “direct assessment of technical skill.”

But first the research needs to be expanded to a larger test group and intelligent virtual-reality machines like the one in Montreal, developed using Canadian research grant money and sold by CAE Healthcare Inc., need to be more widely available.

Dr. Abhaya Kulkarni, pictured, says the latest research using machine-learning is significant because it allows a user’s level of skill to be classified in a relatively objective fashion.

JIM ROSS/THE GLOBE AND MAIL

Effective training that helps future surgeons is good for patients going through a stressful procedure, said Susan Marshall, chief executive officer of the Brain Tumour Foundation of Canada.

“Certainly, the more skills and confidence that the surgeon has, they translate those to the patient,” she said.

Her son Brent was diagnosed with a brain tumour at four years old, and had surgery twice before he died in his 20s. Dr. Del Maestro was his surgeon, and Ms. Marshall was grateful her son recovered both times without complications.

But for other patients, brain surgery can mean long-term deficits in speech and movement. Minimizing surgical mistakes can mean real benefits for patient outcomes, Dr. Del Maestro said.

He envisions surgery following the aviation industry, where pilots enter flight simulators regularly to reaffirm their skills.

“What we’re really trying to do is make people better,” he said.

https://www.theglobeandmail.com/canada/article-how-artificial-intelligence-could-change-how-brain-surgeons-are/