Introduction

Drs. Douglas Flora and Shaalan Beg discuss the use of artificial intelligence in oncology, its potential to revolutionize cancer care, from early detection to precision medicine, and its limitations in some aspects of care.

TRANSCRIPT

Dr. Shaalan Beg: Hello and welcome to the ASCO Daily News Podcast. I'm Dr. Shaalan Beg, your guest host of the podcast today. I'm the vice president of oncology at Science37 and an adjunct associate professor at the UT Southwestern Medical Center in Dallas.

On today's episode, we'll be discussing the use of artificial intelligence in oncology, its potential to revolutionize cancer care from early detection to precision medicine, and we'll also go over limitations in some aspects of care. I'm joined by Dr. Douglas Flora, the executive medical director of oncology services at St. Elizabeth Healthcare in northern Kentucky, and the founding editor-in-chief of AI in Precision Oncology, the first peer-reviewed, academic medical journal dedicated specifically to advancing the applications of AI in oncology. The journal will launch early next year.

You'll find our full disclosures in the transcript of this episode and disclosures of all guests on the podcast are available at asco.org/DNpod. Doug, it's great to have you on the podcast today.

Dr. Douglas Flora: I'm glad to be here. Thanks for having me.

Dr. Shaalan Beg: First of all, Doug, congrats on the upcoming launch of the journal. There has been a lot of excitement on the role of AI in oncology and medicine, and also some concern over ethical implications of some of these applications. So, it's great to have you here to address some of these issues. Can you talk about how you got into this space and what motivated you to pursue this endeavor?

Dr. Douglas Flora: I think, Shaalan, I've embraced my inner nerd. I think that's pretty obvious. This is right along brand for me, along with my love of tech. And so, I started reading about this maybe 5, 6, 7 years ago, and I was struck by how little I understood and how much was going on in our field, and then really accelerated when I read a book that the brilliant Eric Topol wrote in 2019. I don't know if you've seen it, but everything he writes is brilliant. This was called Deep Medicine, and it touched on how we might embrace these new technologies as they're rapidly accelerating to ultimately make our care more human. And that really resonated with me. You know, I've been in clinical practice for almost 20 years now, and the same treadmill many medical oncologists are on as we run from room to room to room and wish we had more time to spend in the depths of the caves with our patients. And this technology has maybe lit me up again in my now 50-year-old age, say, wow, wouldn't it be great if we could use this stuff to provide softer, better, smarter care?

Dr. Shaalan Beg: When I think about different applications in oncology specifically, my mind goes to precision oncology. There are many challenges in the precision oncology space from the discovery of new targets, from finding people to enroll them on clinical trials, ensuring the right person is started on the right treatment at the right time. And we've been talking a lot about and we've been reading and hearing a lot about how artificial intelligence can affect various aspects of the entire spectrum of precision medicine. And I was hoping that you can help our listeners identify which one of those efforts you find are closest to impacting the care that we deliver for our patients come Monday morning in our clinics and which have the highest clinical impact in terms of maturity.

Dr. Douglas Flora: You know, I think the things that are here today, presently, the products that exist, the industry partners that have validated their instruments, it's in 2 things. One is certainly image recognition, right? Pattern doctors like dermatologists and people that read eye grounds and radiologists are seeing increasing levels of accuracy that now are starting to eclipse even specialists in chest radiology and CT or digital pathology with pixelated images now for companies like Path AI and others are publishing peer review data that suggests that the accuracy can be higher than that of a board-certified pathologist. We're all seeing stuff in USA Today and the New York Times about passing medical boards and passing the bar. I think image recognition is actually right here right now. So that's number 1.

Number 2, I think is less sexy, but more important. And that is getting rid of all the rote mechanical mundane tasks that pollute your days as a doc. And I mean specifically time spent on keyboard, pajama time, documenting the vast amounts of material we need for payers and for medical documentation. That can be corrected in hours with the right programming. And so, I think as these large language models start to make their way into clinic, we're going to give doctors back 3, 4, 6 hours a day that they currently spend documenting their care and let them pay attention to their patients again, face to face, eye to eye.

Dr. Shaalan Beg: I love the concept of pajama time. It's sort of become normalized in many folks that the time to do your charting is when you're at home and with your family or in your bedroom in your pajamas, cleaning notes and that's not normal behavior. But it has been normalized in clinical care for many reasons, some necessary and just some not maybe so much. We hear about some of the applications that are coming into electronic medical records. It's been many years since I saw this one demo which one of the vendors had placed where the doctor talks to the patient and then asks the electronic medical record to sum up the visit in a note and then voila, you have a note and you have the orders and you have the billing all tied up. It's been at least 4 years since I've seen that. And I'm not seeing the applications in the clinic or maybe something's turning around the corner because for a lot of people, AI and machine learning was just an idea. It was pie in the sky until chat GPT dropped and everybody got to put their hands on it and see what it can produce. And that's literally scratching the surface of what's possible. So, when you think about giving the doctors their pajama time back, and you think about decision support, trial matching, documentation, which one of those applications are you most excited about as an oncologist?

Dr. Douglas Flora: I'm still in the trenches. I just finished my Wednesday clinic notes Friday afternoon at 4:30 pm, so I think medical documentation is such a burden and it's so tedious and so unnecessary to redouble the efforts again and again to copy a note that four other doctors have already written on rounds It's silly. So, I think that's going to be one of the early salvos that Hospital systems recognize because there's a higher ROI if you can give 400 doctors back two hours a day. It's also satisfying because the notes will be better. The notes will be carefully curated. They may bring in order sets for the MRI with gadolinium that you forgot you wanted to order; the digital personal assistant will get that. It will set a reminder on your calendar to call the patient back with their test results. It will order the next set of labs, and you're going in the next room, and you're going to be watching that patient in the room. And I've talked to other colleagues about this earlier today. You'll be able to see the daughter getting hives because you're watching her or the look that fleets across the husband's face when you go a little bit too far and you go out too much information when they're not quite ready for that. And I think that's the art of oncology that we're missing when we're flying in a room, and we've got our face on the screen and a keyboard, and we're buried in our own task and we're not there to be present for our patients. So, I'm hopeful that that's going be one of the easy and early wins for oncologists.

Dr. Shaalan Beg: Fantastic. And when we think about the spectrum of cancer care for the people who we care for, a lot happens before they walk into their medical oncologist's office in terms of early identification of cancer, just the diagnosis of cancer, the challenges around tissue acquisition, imaging acquisition. You mentioned a couple of the tools around radiomics, which are being implemented right now. Again, same question: Separate fact from fiction, which ones are we going to see in 2023 or 2024 in the clinical practice that we have? We've been hearing that pathologists and radiologists are going to be out of their jobs if AI takes off, right? Of course, that is a lot of hyperbole there. But how do you view that space and how do you see it impacting the overall burden of care that people receive, and the burden of care that physicians are experiencing?

Dr. Douglas Flora: I'm an eternal optimist, almost infuriating optimist to my partners and colleagues. So, I'm going to lean into this and say, burdens are going be reduced all over the place. We're going to have personal digital navigators to help our patients from the first touch so that they're going to have honest and empathetic questions answered within an hour of diagnosis. The information that they're going have at their fingertips with Chatbot 4 or Med-PALM 2 with Google that's about to be released as a medical generative AI. These are going to give sensitive and empathetic answers that don't put our patients on the cliff, you know, that they're falling off waiting for a doctor's visit 10 days down the road. So, I think the emotional burdens will be improved with better access to better information. I think that the physicians will also have access to that, giving us reassurance that we're going down the right path in terms of really complicated patients taking very, very large datasets and saying a digital twin of this patient would have been more successful with this approach and those sorts of things. And those are probably 3 to 5 years down the road but being tested heavily right now in academic settings with good data coming.

Dr. Shaalan Beg: Robotic empathy sounds like an oxymoron.

Dr. Douglas Flora: Yeah, look at the published studies.

Dr. Shaalan Beg: We've all seen the data on how a chatbot can outperform physicians in terms of empathy. I really find that to be hard to stomach. Help me out.

Dr. Douglas Flora: Yeah, we say that, and we say that to be provocative, but no, there's no substitute for a clinician laying a hand on a patient. We talked about how you need to see that fleeting glance or the hives on the daughter's chest and that you've gone too far and shared too much too soon before that family is ready for it. I have no doubt in my mind, these tools can make us more efficient at our care, but don't get me wrong. There's no chance that these will replace us in the room, giving a hug to a patient or a scared daughter. They're going to remember every word you say; I just want it to be the right words delivered carefully and I don't want us to rush it. So ultimately, as we make our care more human, these tools might actually give us time back in the room to repair that doctor-patient relationship that's been so transactional for the last 4 or 5 or 10 years. And my hope is, we're going to go back to doing what we went into oncology to do, to care deeply about the patients in our care and let the computers handle the rote mechanical stuff; let me be the doctor again and deserve that patient's attention and give it right back in return.

Dr. Shaalan Beg: And I think we're hearing a lot of themes in terms of AI helping the existing clinical enterprise and helping make that better. And it's not your deep blue versus Kasparov, one person is going to win. It's the co-pilot. It's reducing burden. It's making the work more meaningful so that the actual time that's spent with our patients is more meaningful and hopefully can help us make deeper connections.

Let's talk about challenges. What are some of the challenges that worry you? There've been many innovations that have come and gone, and health systems and hospitals have resisted change. And we all remember saying during COVID that we're never going to go back to the old ways. And here we are in 2023 and we are back to the old ways for a lot of things. So, what are the major limitations of AI, even at its... peak success that you see, which our listeners should be aware of, and which may worry you at times.

Dr. Douglas Flora: Well, you've actually spoken to why I started this journal. I want to make sure that clinicians are guiding some of those conversations to make sure that guardrails are up so that we're ethical and we are making sure that we are policing bias. It's no secret now you've seen these things – a lot of language models, a lot of the deep learning was programmed by people that look like me and did not include things that were culturally competent. You can look at data that's been published on Amazon and facial recognition software for Facebook and Instagram and others. And they can identify me out of a crowd as a middle-aged white guy, but 60% of the time they will not recognize Oprah Winfrey or Serena Williams or Michelle Obama. I mean, iconic global icons. And with darker skin, with darker features, with different facial features than my white Caucasian, Eurocentric features, these recognition softwares are not as good. And I'm worried about that for clinical trial selection and screening for that. I'm really, really worried about building databases that don't represent the patients in our charge. So bias is a big deal and that's got to be transparent. That's got to be published how you arrived at this decision. And so that would be number 1. Number 2 is probably that we don't have as much. visibility to how decisions are made, this so-called black box in AI. And that's vexing for doctors, especially conservative oncologists that need 3 published randomized phase 3, blinded, placebo-controlled trials before we move an inch. So, there must be more transparency. And that again is in publications, it's in peer review. They say we need real scientific rigor and not to belabor this, but our industry partners are well ahead of us. We're not generally inclined to believe them until we see it because I've got 150 AI companies coming to my hospital system as vendors some of them are worthy great partners and some of them are a little bit over their skis and selling more than they can actually deliver yet. So, I'd like to give that an opportunity to see the papers. There's about 300 produced a day in AI in medicine. Let's give them a forum and we'll duke it out with letters of the editor and careful review.

Dr. Shaalan Beg: I will say Doug, it is becoming hard to separate fact from fiction. There is so much information which is coming across us in medical journals and through our email, through our professional social media accounts that I sometimes worry that people will just start tuning it all out because they can't separate the high impact discoveries from the more pie in the sky ideas. So, tell us more about how we got here and how you see this curve of enthusiasm shifting maybe in the next 6 months or 1 year.

Dr. Douglas Flora: Yeah, it's a great question. And it's rapidly accelerating, isn't it? We can't escape this. It's entering our hourly lives, much like the iPhone did before, or me having to switch from my BlackBerry to a smartphone that didn't have buttons. I felt like I was adapting. And maybe this is what people felt like when Henry Ford was out there, and all the buggy drivers were getting fired. The reality is it's here and it was here 6 months ago. And maybe we're feeling that urgency and maybe it's starting to catch on in general society because the advent of generative AI is easier to understand. These aren't complicated mathematical models with stacking diagrams and high-tech stuff that's just happening in Palo Alto. It's Siri, it's Cortana. It's my Google digital assistant notifying me that it's time to get on for my next meeting. And those things have been infiltrating our daily lives and our minds quietly for some time. About November 30th when chatbot GPT-3 came out from OpenAI and we started toying with it, you started to see the power. It can be creative, it can be funny, it can articulate your thoughts better than you can articulate them on paper immediately. English students have figured it out. People in marketing and writing legal briefs have figured it out and it's coming to medicine now. It is actually here, and this might be one instance where I think the hype is legit. and these tools will probably reshape our lives.

There have been some estimates by Accenture that 70% of jobs in medicine are going to be altered irretrievably by generative AI. And so, I think it's incumbent upon those of us that are leaders in healthcare systems to at least assemble the team that can help make sense and separate, like you said, the signal from the noise. I know we're doing that here at St. Elizabeth Healthcare. We've got a whole team being formed around this. We have 5 or 6 different products we bought. that we're using to help read mammograms and read lung nodules and read urinalyses, etc. You need a construct to do that appropriately. You need a team of people that are well read and well-studied and able to separate that fact from fiction. I think we're all going to have to work towards that in the next 6 to 12 months.

Dr. Shaalan Beg: Tell me about that construct. How did you, what is the framework that you use to evaluate opportunities as they come through the door?

Dr. Douglas Flora: It's something I think we're all struggling with. As I mentioned, we've got all of these fantastic industry partners, but you can't buy 200 products off the shelf as Epic add-ons as third-party software to solve 200 problems. So, it's interesting, you've just said this. I just shared a piece on LinkedIn that I loved. "Don't pave the cow's path." It's a really thoughtful thing to say, "Before you build an AI solution, let's make sure we're solving the correct problem." And the author of that piece on Substack said: Let's not use AI to figure out how to have more efficient meetings by capturing our minutes and transcribing them immediately. Let's first assess how many of these meetings are absolutely necessary. What's the real job to be done and why would you have 50% of your leadership team in meetings all day long and capture those in yet another form? Let's take a look first at the structure around the meetings and say, are these necessary in 2023 and are these productive?

So, my thought would be as we're starting this. We're going to get other smart people who are well-read, who are studying, who are listening to experts that do it six months ahead of us, and really doing a careful contemplative look at this as a team before we dive in with both feet. And there are absolutely tools that are going to be useful, but I think the idea, how do we figure this out without having 200 members of my medical staff coming to me saying, you've got to purchase all 200 of these products, and have a way to vet them scientifically with the same rigor you would for a journal before you put out that kind of outsource.

Dr. Shaalan Beg: Doug, thanks for coming on the podcast today and sharing your valuable insights with us on the ASCO Daily News Podcast. We'll be looking out for your journal, AI in Precision Oncology, early next year. Tell our listeners where they can learn more about your journal.

Dr. Douglas Flora: I really appreciate you guys having me. I love this topic, obviously, I'm excited about it. So, this journal will be ready for a launch in early October in a preview. And then our premier issue will come out in January. We're about to invite manuscripts in mid-August. I guess parties that are interested right now go to Doug Flora's LinkedIn page because that's where I'm sharing most of this and I'll put links in there that will lead you to Liebert's site and our formal page and I think we can probably put it in the transcript here for interested parties.

Dr. Shaalan Beg: Wonderful. Thank you very much and thank you to our listeners for your time today. Finally, if you have any insights on if you value the insights a little. And thank you to our listeners for your time today. Finally, if you value the insights that you hear on the podcast, please take a moment to rate, review and subscribe wherever you get your podcast.

Disclaimer:

The purpose of this podcast is to educate and to inform. This is not a substitute for professional medical care and is not intended for use in the diagnosis or treatment of individual conditions.

Guests on this podcast express their own opinions, experiences, and conclusions. Guest statements on the podcast do not express the opinions of ASCO. The mention of any product, service, organization, activity, or therapy should not be construed as an ASCO endorsement.

Find out more about today's speakers:

Dr. Shaalan Beg

@ShaalanBeg

Dr. Douglas Flora

St. Elizabeth Healthcare

Follow ASCO on social media:

@ASCO on Twitter

ASCO on Facebook

ASCO on LinkedIn

Disclosures:

Dr. Shaalan Beg:

Employment: Science 37

Consulting or Advisory Role: Ipsen, Array BioPharma, AstraZeneca/MedImmune, Cancer Commons, Legend Biotech, Foundation Medicine

Research Funding (Inst.): Bristol-Myers Squibb, AstraZeneca/MedImmune, Merck Serono, Five Prime Therapeutics, MedImmune, Genentech, Immunesensor, Tolero Pharmaceuticals

Dr. Douglas Flora:

Honoraria: Flatiron Health