Toronto doctor developed an AI to reduce surgical complications


Dr Amin Madani, a general surgeon in the Sprott Department of Surgery at the University Health Network (UHN) in Toronto, is trying to reduce the risks of surgery by developing an AI system to assist surgeons.

To begin with, Dr Madani was a big sceptic of the potential of artificial intelligence to improve the rate of successful surgery. He started his research on enhancing surgery outcomes by trying to understand the thought process and techniques practised by elite surgeons. However, the difficulty lay in making the best use of the knowledge gained through this research. Some data and computer scientists suggested he build an artificial intelligence model that can mimic the minds of these highly skilled surgeons.

Despite being a sceptic, he collaborated with data scientists and built a prototype that uses computer vision to identify real-time areas of an organ that are safe to dissect and those regions that are dangerous. The technology, still in its early stages, is ready for application in gallbladder surgery or specifically for the procedure of laparoscopic cholecystectomy (LC).

How does the technology work?

LC typically occurs by making incisions in the abdomen, inserting a camera to see the inside and safely removing the gallbladder. Madani’s AI system gives a real-time coloured projection of the video captured by the camera. Any gallbladder region depicted in green is safe to dissect, while the surgeon should not touch any area in red. The AI system was developed using 290 videos of gallbladder surgery from 37 different countries, 135 different institutions and 153 expert surgeons. The system starts figuring out patterns after analysing the data and becomes capable of making independent decisions. The design of the prototype is detailed in a study in the journal Annals of Surgery.

This novel system recognised safe (Go) and dangerous (No-go) zones of dissection with greater than 90% accuracy, which was determined by cross-checking the system’s predictions against annotations by expert surgeons. The main challenge of implementing this system is the lack of available data for most major surgeries. Operations that are not as common as gallbladder removal do not have enough available datasets in the form of surgery videos to train the system. Madani is well aware of this problem and is holding talks with health experts to create a global repository of surgical videos.


Using CNN for real-time assistance during laparoscopic cholecystectomy

Future of surgery

Dr Amin Madani is a part of a community of researchers and healthcare professionals trying to integrate powerful technologies such as AI to improve medical care. Performing surgery requires extreme skill, focus and precision; researchers hope to develop AI systems to train and assist surgeons. Surgery simulators are already being used for training purposes, and augmented reality is helping surgeons.

The AI system developed by Madani will provide real-time assistance to surgeons and help them make the correct decisions. When a global repository of surgical videos is available for training this AI model, it will have much greater relevance as it will be applicable for different surgeries and act as “a second pair of eyes” to assist doctors in the operation room.