The AIMed Surgery virtual conference took place on 23 September. Similar to the AIMed ICU virtual conference took place the day before, the half-day long event consisted of four main panel sessions on the topics of data, machine and deep learning; natural language processing (NLP) and workflow, and general issues in artificial intelligence (AI) and surgery. There was also a real-time application of AI session to provide a step by step guide on the thought processes needed to solve an actual clinical problem as well as a journal club to review the top three AI in Surgery paper published last year.
Develop solutions that target specific challenges
In one of the sessions, Dr. Arlen Meyers, President and Chief Executive Officer of the Society of Physician Entrepreneurs (SoPE) and Emeritus Professor at the University of Colorado School of Medicine asked fellow panelists to describe the biggest obstacle they thought to prevent the widespread dissemination of AI in the field. Dr. Bethany Slater, Assistant Professor of Surgery at the University of Chicago’s answer was to connect people who know AI and have the skillsets to surgeons. She said the two groups shared different languages and she has tried to pair them in a recent project but was unsuccessful. Thus, she hopes an initiative like AIMed will help to bridge the gap.
For Dr. Max R. Langham Jr., Professor of Surgery and Vice Chairman of Department of Surgery atThe University of Tennessee Health Science Center, the barrier lies in data. He said most data sources are incomplete and contain significant bias – designed to be or unintentional. Quality, amount, and accuracy of data are super important particularly when it comes to adverse events like death after surgery, which is rare. This rarity will get data “muddy quickly”, as such, he wishes to learn more about various AI techniques to overcome the challenge.
Dr. Baran Tokar, Professor of Pediatric Surgery and Pediatric Urology at the Eskisehir Osmangazi University in Turkey expressed his concerns on regulations and ethical issues. He thought they were the main culprits that prevent surgeons from getting enough data. At the same time, he also warned the possibility of generating AI models with “very high accuracy but low clinical outcomes” and urged a closer partnership between data scientists and surgeons to understand one another’s needs.
Ensuring the technologies support or improve current practices
Dr. Prakash Gatta, Esophageal Surgeon at the MultiCare Health System said AI has a greater penetration in general surgery as compared to pediatric surgery especially in the last couple of years. He cited surgical delivery using robots as an example. As robot is a device and leaves behind digital trails or imprints (i.e., the movement made and instruments handled during surgeries) which human are able to re-create virtually. They can be used to fill in the knowledge gaps that surgeons have as a result of forming the surgical experiences based solely on metrices and improve their insights overtime.
“It will be great to have a robot which tells me at the end of the operation, you did a poor job because you used this instrument 60% of the time longer than you usually do. If I am early in my practice, the robot will know what is considered an efficient operation and used that to judge my performance. So, I will have a score. It’s like continuing your residency but instead of having a human expert, a robot or a technology will be telling me where I can do better,” Dr. Gatta says.
In short, US is likely to face a shortage of 50,000 surgeons in the next 20 to 30 years. Hence, Dr. Gatta believes there is a need to make surgical training more virtual, more technical, ideally with augmented virtual reality as the mechanism. Dr. Tokar echoed the comment. He thought about a kind of cognitive assistance system providing guidance in terms of recommendations of tools and pointing out errors. He and his research team had embarked on a deep learning project, by dissecting videos of surgeons in action and study those pictures where a mistake was made. In another words, looking at visual data like the way a machine does and from there, derive learning points where surgeons can do better.
Many answers will come down to simulations
On the other hand, Dr. Slater wishes AI can enhance simulation so that surgeons will have more hands-on practices. She said as a pediatric surgeon, some cases only appear once a year. So, technology can definitely come in to provide training and ensure surgeons have the skillsets to tackle cases that they have not met before or have not done in a while. Dr. Anthony Chang, AIMed Founder and Chief AI Officer at the Children’s Hospital of Orange County (CHOC) also suggested the possibility of a virtual twin situation whereby surgeons will be trained on a virtual replica of a patient rather than a real patient and this is a new dimension that surgeons should prepare themselves for.
Overall, speakers believe there is a huge opportunity to go forward to improve simulation and its role in surgical training. As mentioned by Dr. Gatta, simulation will come in handy for surgeons who have very busy clinical practice and are keen on learning or utilizing robotic platform but do not have the time to receive training or do not wish to make a loss of revenue from the learning curve.
Nevertheless, Dr. Arlen thought these types of assistance and training may lead one to think about credential and licentiateship. He said medical practitioners undergo periodic training to acquire new skills or knowhow of new tools and how to credential AI as an integrated surgical technology may create yet another challenge. With that, Dr. Langham explained AI can be viewed as a tool in credential. One can envision an AI model and its portfolio of cases combined with assessment on whether it has provided good risk stratification, user input, patient feedback, how well the surgeons are doing with the tool and so on, to be submitted to a credential body.
The AIMed Surgery virtual conference is available on demand here.