Advertisement


Andrew Srisuwananukorn, MD, and Alexander T. Pearson, MD, PhD, on Artificial Intelligence in the Clinic: Understanding How to Use This 21st Century Tool

2024 ASCO Annual Meeting

Advertisement

Andrew Srisuwananukorn, MD, of The Ohio State University, and Alexander T. Pearson, MD, PhD, of the University of Chicago, discuss the use of artificial intelligence (AI) in the clinic, its potential benefits in diagnosis and treatment, resources available to help physicians learn more about AI, and what’s coming for the next generation of medical school students.



Transcript

Disclaimer: This video transcript has not been proofread or edited and may contain errors.
Thank you for joining me today, Alex. And I wanted to ask you a couple questions about a mutual interest of ours, and that's artificial intelligence, or AI, in medicine and cancer care specifically. You're a physician as well, and an AI researcher. How do you think about AI or how do you think that physicians should be thinking about AI? Definitely. AI is a tool, or a group of tools, and I think of AI as a tool like any other at our disposal. You mean like other statistical tools? Or even other biomarkers. Any other clinical tool. Applied to the right dataset? AI has the ability to process data in a way to give us the most extractable information from any form of clinical information that we're collecting on behalf of our patients. Right, right. Oh, that's very interesting. I get a lot of questions in my practice about what can AI do? What are the types of questions do you think are really important in cancer care? I guess I think about it in terms of what it can do now and what we think it might be able to accomplish. Those are two separate components. I think now we're learning that AI is a series of methods that are able to extract all kinds of inferential data from different forms of information that we're collecting on behalf of our patients already. For example, understanding the molecular state in a person's colorectal cancer directly from automatic evaluation of the pathology slide, saving them the need for getting a specific additional molecular test. I mean, that might save our patients time and cost in order to get to a precision cancer decision point really fast. Interesting, interesting. So what I'm hearing is that AI can be very flexible and very cost-efficient and potentially efficacy everywhere. Yeah. I mean, it's not going to work, it's not magic. It's not going to- It's fair. It's a system of building connections between outcomes and data, for example. But if those connections don't exist, AI is not going to invent it. Great, great, great. So then which domains in medicine do you think AI will be adopted in? Yeah, I think it's going to touch every domain of medicine. Interesting. Because it's a family of methods, but it's a family of methods that can process voice information, can process text information, can process imaging information both in radiographic images or in scope images or in pathology images. All of these are submodules or specific types of AI. Right. I think it's going to be comprehensive in terms of developed exacting tools that have very specific uses that touch all aspects of our care for our patients. Right, right. And I think it's going to come into many of our practices. We see these generative models in real world practices with the large language models and image generation. So as these algorithms come, who do you think will be adopting these types of models and algorithm, and really, what do you think will be influencing where they lie in our practice? Yeah, it's an excellent question and I think it's something that we should be keeping in mind as providers, because I think you and I think about AI as a series of tools that can help our patients. Right. It can accomplish a task with less cost in faster time. But I think that we need to also be mindful that if AI could accomplish or seek to accomplish tasks that other stakeholders want to accomplish, for example, that the cost component if implemented in the wrong way, if we're pursuing only a cost basis, it might lead to worse care on behalf of patients if physicians aren't involved in some of that decision making. It's so important. Right, right, right. And that's my biggest fear too, is that it may lead to, in certain situations, harder situations or worse decision making. Yeah. You can see it changing our practice if we are left to only make decisions or only have patients that AI has been unable to help or assist with. It's going to be a very different experience both for providers and for patients. Right, right. For example, my patient volume may be 20 patients a day right now without AI but if AI only triages us a certain patient population, do you think that'll be more challenging for physicians? Yeah. I could imagine a scenario where with the number of assistive devices for a practice that you're responsible for multiple times more patients and that those patients are interacting with chatbots or are being evaluated by an automatic scan device, for example. Right. And so the number of patients that we're responsible for as providers actually increases over time. That is dependent on being able to trust these algorithms that we're working with. Right, right, right. So I feel it's very important for us physicians to be really involved in these processes to make sure that we know how to implement these algorithms in our practice and what's the best way to do that? I totally agree. So my last question for you is there's so much interest in AI, AI in medicine, AI in cancer. Could you offer some advice to maybe trainees or even the field in general about how do we learn about this? How do we learn about AI? How do we do better? Yeah. I think all of the main stakeholder groups, including ASCO, are thinking a lot about how do we provide educational resources to the community at large, both patients and providers. I think we're going to see a lot more rich information coming. I think in the current best practice for providers currently is to look in the medical literature in that, for example, journals like NEJM AI and some of the big scientific groups all have a lot of resources available to learn about very specific nuanced subcomponents of artificial intelligence. I think what it's going to be, I mean you and I have talked about this, in that neither of us had AI in our medical school curriculum, but it's going to be there in 10 years. I think it's going to come. Positive predictive value is going to be side by side with area under the receiver operator curve. The next generation of medical education is really going to involve AI. I 100% agree, and I think it'll be great when that's boiled into the standard of practice and we're testing ourselves on our knowledge, but I think we will be the generation that ensures the next generation gets that education. Perfect. Perfect. Thank you so much, Alex. I really appreciate- This has been great.

Related Videos

Prostate Cancer
Genomics/Genetics

Alicia Morgans, MD, MPH, and Susan Halabi, PhD, on Prostate Cancer: New Findings on Classifying Patients Into Risk Groups

Alicia Morgans, MD, MPH, of Dana-Farber Cancer Institute, and Susan Halabi, PhD, of the Duke Cancer Institute and Duke University School of Medicine, discuss a clinical-genetic model that identified novel circulating tumor DNA alterations that are prognostic of overall survival and may help to classify patients with metastatic castration-resistant prostate cancer into risk groups useful for selecting trial participants (Abstract 5007).

Skin Cancer

Omid Hamid, MD, on Cutaneous Melanoma: Update on a Bispecific Protein Under Study

Omid Hamid, MD, of The Angeles Clinic and Research Institute, a Cedars-Sinai affiliate, discusses updated data on IMC-F106C, a novel bispecific protein that, in a phase I safety and efficacy study, exhibited clinical activity in patients with unresectable or metastatic cutaneous melanoma who were pretreated with immune checkpoint inhibitors. A phase III trial of IMC-F106C with nivolumab in the first-line setting of metastatic disease has been initiated (NCT06112314; Abstract 9507).

Lymphoma

Peter Riedell, MD, on DLBCL: Expert Commentary on the DEB Study

Peter Riedell, MD, of The University of Chicago, discusses phase III results on the use of tucidinostat plus R-CHOP in patients with previously untreated diffuse large B-cell lymphoma (DLBCL) with double expression of MYC and BCL2. The regimen appeared to improve event-free survival and complete response rates vs R-CHOP in the front-line setting. As this is an interim analysis, longer-term follow-up will be needed to better understand its impact, says Dr. Riedell.

Prostate Cancer

Christos Kyriakopoulos, MD, on Prostate Cancer: CHAARTED2 Trial Results on Cabazitaxel and Abiraterone

Christos Kyriakopoulos, MD, of the University of Wisconsin Carbone Cancer Center, discusses data suggesting that adding cabazitaxel to abiraterone and prednisone improves progression-free survival in patients with metastatic castration-resistant prostate cancer who previously received chemohormonal therapy with docetaxel for hormone-sensitive disease compared with abiraterone plus prednisone alone (Abstract LBA5000).

Lung Cancer

Tony S.K. Mok, MD, on NSCLC: Adagrasib vs Docetaxel in KRAS G12C–Mutated Disease

Tony S.K. Mok, MD, of The Chinese University of Hong Kong, discusses phase III findings from the KRYSTAL-12 study, which showed that adagrasib improved progression-free survival and overall response rate over docetaxel in patients with locally advanced or metastatic non–small cell lung cancer harboring a KRAS G12C mutation who had previously received a platinum-based chemotherapy with anti–PD-(L)1 treatment.

Advertisement

Advertisement




Advertisement