AI for Mohs Surgery

In this exclusive interview, Dr Vidal discusses her session, “Cutting Edge Technology in Dermatologic Surgery,” at the 2023 ACMS Annual Meeting.

Nahid Y. Vidal, MD, FACMS, is the chair of the division of dermatologic surgery at the Mayo Clinic in Rochester, MN. She also serves as principal investigator for research focused on artificial intelligence (AI) and innovation in cutaneous surgery and oncology.

The Dermatologist: Can you give us a recap of what was covered during your session at the 2023 ACMS Annual Meeting?

Dr Vidal: Our session was moderated by Dr Clark Otley and Dr Kishwer Nehal. Dr Kalee Shah introduced machine learning and AI in general. We also had a keynote speaker the day before who had primed our audience in preparation for what I spoke about, which was AI progress, specifically in Mohs surgery.

My talk was broken down into summarizing current literature and progress in AI in Mohs surgery, and then exploring the question of whether we think as a specialty AI will replace us or a component of us. And then finally discussing the potential future for AI and being able to help us solve problems and serve patients while keeping patient equity in mind, including the regulatory relevance. We ended with Dr Nehal giving us a nice summary and more detailed exploration into microscopy devices and the progress they have made.

The Dermatologist: How are the AI results applicable in clinical decision-making?

Dr Vidal: I broke my talk down into preoperative, intraoperative, and postoperative applications. Preoperative, I think there is a lot that can be done from an administrative standpoint to help the administrative burden overall. One of the examples that I gave was a couple of international studies about triaging cases or how to optimally schedule patients in the right place, looking at their pathology reports and tumor characteristics and patient demographics. I think there is a lot of growth in that area for clinical applications. How do we schedule patients so they are not only being scheduled appropriately, but also so we can prime them from an education standpoint and help better prepare them for their surgeries?

 We also spoke about making progress before they even come to us in Mohs surgery. Right now the standard of care is the patient has a lesion; they get a biopsy; they wait however many days or weeks for the pathology report, which then tells them whether they have cancer and what type; and then from there their dermatologist will make a recommendation for surgery or other treatments. The hard part about biopsies is that not all the margin is going to be evaluated. Sometimes the biopsy has taken all the skin cancer out already and sometimes not. We do not have a reliable way of knowing that and so, for the most part, we treat all skin cancers of special sites with Mohs surgeries.

A very common question that patients have is, this has healed so well since the biopsy, do I really need this surgery? And there is quite a bit of literature showing that anywhere between 30% and 70% of the time, there is residual skin cancer left behind, even when clinically we cannot see it. And so we typically proceed with doing Mohs surgery. But with these preoperative bedside diagnostics that are being enabled with AI, I think we are getting to a place where we could potentially look at those margins just from the biopsy specimen and know right then and there whether they even need surgery or whether the biopsy took care of it. I think we will have a lot less unnecessary surgery being performed.

In terms of intraoperative, there are many areas, I think the way we currently do our precise mapping for Mohs surgery that allows us to get a very high cure rate, 98% to 99%, relies on a surgeon doing grossing of tissue, which means separating it and mapping along precisely on a piece of paper or sometimes digitally. This a manual process to freeze the tissue, cut it into thin sections that are microns thick, put it onto slides, go through a staining process, create those slides, and take a look at them under the microscope. I can see a time in our future where everything is being digitally mapped in real time by AI onto a patient's wound defect or a photo of it. And that way we can be even more precise with our mapping techniques. The other thing I talked about is just the way that we look at tissue in general. I talked about the potential to make our lab process more efficient or even in some cases bypass it altogether if we can have a 3D microscope enabled with AI to allow us to look at all those margins without even going through the lab staining process at all.

Finally, for postoperative, I think there are a lot of areas where we can make great progress, mostly in the prognostication and treatments that right now are based on staging research, as well as retrospective prospective data. I can see a time in the future where we can stratify the risk of a patient having things like metastasis or local recurrence, and then from that stratification being able to provide recommendations for how often they should be screened for skin cancer or whether they need radiation just based on the individualized treatment plan.

And then we chatted a little bit about what to expect and how soon during the question-and-answer portion of the session. I think we all felt like right now it is not happening in our everyday life, primarily because we do have a regulatory cycle where the FDA is there to ensure that we are not just creating amazing algorithms, but we are making sure they are based off diverse data sets, they are generalizable, they have been validated, and ultimately we have patient safety at the center. I think that will be part of what determines how fast AI becomes applicable in our daily lives. And then the other thing I think will determine that is just the infrastructure and societal acceptance of us using AI on a day-to-day basis.

The Dermatologist: Can you go over the key points from your session?

Dr Vidal: My take-home message in general was that AI is happening quickly. The pace of progress is fast. And unlike a lot of areas of clinical research in medicine, publications are not always a measure of how much progress is being made. We talked a little bit about large language models like ChatGPT already showing how impressive they are, but then we also spoke about the dangers and that we do have to slow down. One of the main take-home points of my talk was that Mohs surgeons need a voice and they need a seat at the table, not only from the design and framework development, but at every step of the way. I am not overly concerned with AI replacing me, but I am concerned about AI being developed at such a fast pace that we become irresponsible with it or confused with who is taking liability in the way that we use it. Another take-home point was that we just have to make sure it is equitable, that it is ultimately serving patients.

And then, finally, the other point that I made was when we are trying to think about what we can do with AI and how we can all work together and collaborate, it is helpful to think about our job and how we can better serve patients. For example, what is burdensome, what is impossible for us when it comes to serving our patients? What is difficult for us, finding high-risk features on pathology? All those things I think will help us, guide us in the right direction so the algorithms being made are there to ultimately improve patient care and to have their safety and their outcomes in mind.

The Dermatologist: What additional insights would you like to share with your colleagues regarding AI for Mohs surgery?

Dr Vidal: I am going to make a big call for collaboration. I think there are multiple people, not just within Mohs surgery, but also an industry that does not include our voice where people are developing in parallel. There is more power and more integrity of data if we are working together. I would love to collaborate with other people who are interested in this area, who are patient-centric and have that protective nature toward patients. I think it is an area where anyone who has interest can learn. If anyone is intimidated by AI, there are so many available resources to learn from the beginning, from scratch, that are free resources online. I think that it is welcome to anyone who has a passion for development and an innovative mindset.

Reference:
Vidal NY. Cutting edge technology in dermatologic surgery. Presented at: American College of Mohs Surgery (ACMS) Annual Meeting; May 4–7, 2023; Seattle, WA.

Listen to Dr Vidal's interview podcast!