As Artificial Intelligence (AI) finds its way into all aspects of daily life, soon it could be an important component of scientific research, patient care, and the ins and outs of clinical practice. On November 8 and 9, the Endocrine Society will host the AI in Healthcare Virtual Summit that will cover a variety of topics that will inform, educate, and, most importantly, prepare today’s practicing endocrinologist for tomorrow.
In 2024, artificial intelligence (AI) is somehow involved in almost every facet of life, but people still seem murky on exactly how it works or what it even is. The problem is that most people – even healthcare professionals – get their news about AI when someone generates a fake picture of a celebrity or uses AI to write an academic paper that turns out to be wildly inaccurate.
On November 8 and 9, the Endocrine Society is hosting the AI in Healthcare Virtual Summit, a two-day virtual event designed to inform providers, healthcare professionals, researchers, technologists, industry stakeholders, and educators on the capabilities of artificial intelligence in the healthcare field.
“A lot of physicians hear about AI in healthcare from the news or vendors or commercials,” says Jeffrey Moon, MD, MPH, assistant chief medical officer at the University of Pennsylvania in Philadelphia and one of the presenters at the summit. “They don’t hear it from their fellow doctors and healthcare providers who are unbiased and have a scientific background. That’s my role and I will be able to provide that to like-minded colleagues, an objective assessment on where we are today and where we’ll be in the future.”
The AI Summit agenda is robust and diverse, from using the technology in diabetes treatment to diagnosing thyroid cancer to using AI to guide decision making. Endocrine News caught up with a few of the presenters to find out a little about their presentations and their thoughts on the current state of AI in healthcare.
AI and Thyroid Cancer
Nikita Pozdeyev, MD, PhD, is assistant professor of Biomedical Informatics at the University of Colorado Anchutz School of Medicine and an endocrinologist specializing in the diagnosis and treatment of thyroid diseases, including thyroid cancer. He runs a “dry” laboratory focusing on data analysis.
“We believe that learning from the data responsibly can improve clinical care for patients with thyroid cancer and other thyroid diseases,” Pozdeyev says. “My interest in using AI methods has grown since the invention of convolutional neural network technology, when computers began outperforming humans in comprehending visual data, including medical images.”
Pozdevev’s talk, “Artificial Intelligence and Statistical Genetics for Diagnosing Thyroid Cancer,” will, as the title suggests, focus on AI and machine learning for thyroid cancer diagnosis. He says that deciding which nodules need biopsy is complex and subjective.
“Because of this difficulty, clinicians biopsy too many thyroid nodules, which increases healthcare costs and frequently leads to additional testing, diagnostic surgery, and patient stress,” Pozdeyev says. “AI can streamline the interpretation of thyroid ultrasound images, but it has limitations that I will highlight in the talk. We work to augment AI predictions using alternative thyroid cancer risk assessments, such as by quantifying the inherited risk of developing the disease.”
“We believe that learning from the data responsibly can improve clinical care for patients with thyroid cancer and other thyroid diseases. My interest in using AI methods has grown since the invention of convolutional neural network technology, when computers began outperforming humans in comprehending visual data, including medical images.” — Nikita Pozdeyev, MD, PhD, assistant professor, Biomedical Informatics, University of Colorado Anchutz School of Medicine, Aurora, Colorado
Pozdeyev says he will make his presentation as entertaining as possible but adds that the main takeaway will be that “AI is a machine-learning tool that is very powerful. It is now able to perform tasks previously thought to be doable by humans only. However, AI does not reason or feel, and we should use it to augment but not replace clinical providers.”
Pozdeyev says that we are at the “Peak of Inflated Expectations” of the Gartner hype cycle. “I do not mean this in a negative sense; it is a necessary step that will ultimately result in the safe and productive use of AI in medicine,” he says.
AI in Decision Making
Wuraola Oyewusi is a data scientist and AI technical instructor at LinkedIn Learning previously held roles in AI Research as a researcher (data science and data curation) at Imperial College London and led Research and Innovation at Data Science Nigeria.
“I started my career as a pharmacist before transitioning into Data Science and AI,” Oyewusi says. “My interest in AI was somewhat incidental. I came across a job description related to analytics and thought I could do most of the tasks listed, except for SQL (a language used to retrieve data from databases in the way you need). So, I decided to take an online course on SQL, and it made sense to me.”
Oyewusi says that from there, she discovered data science and AI, and she found them fascinating. “This led me to dive deeper into learning everything I needed, from programming to linear algebra and machine learning (I treated it like a school schedule),” she says. “Once I had a solid grasp of these concepts, I found it easy to apply them within the context of my healthcare background.”
“The concerns [about AI] are valid, and it’s important for people to ask questions. Like with any other tool, it’s our responsibility as professionals to become informed and engage with AI, rather than viewing it as some kind of magic.” — Wuraola Oyewusi, data scientist, AI technical instructor, LinkedIn Learning
Oyewusi’s presentation, “AI for Healthcare in Action: What You Need to Know as a Decision Maker” will home in on the fact that AI is an important technology that everyone will be using, and that it’s not some sort of fantastical nightmare. “Now that AI is here, I understand that not everyone will become a coder,” she says. “However, we’ll focus on the key things you need to know in order to make informed decisions about AI systems, without resorting to fear mongering.
“The concerns [about AI] are valid, and it’s important for people to ask questions,” Oyewusi continues. “Like with any other tool, it’s our responsibility as professionals to become informed and engage with AI, rather than viewing it as some kind of magic.”
AI in Clinical Practice
And speaking of concerns, Moon agrees that the hesitation some might feel about AI is well founded, but he says he has not seen anything dangerous regarding safety or hallucinations (when the AI produces inaccurate information that’s presented as fact) or a perpetuation of bias. “These are big things we have been watching like a hawk for,” he says.
But while AI hasn’t been dangerous, it hasn’t been all that great either, Moon says. “If you ask it to help with patient messages that they send to their doctors, the responses are B-plus at best,” he says. “They’re not dangerously wrong, but they’re not that good. If the goal is to help doctors with all the patient messages, it hasn’t really done that.”
Moon is an emergency physician and has a particular interest in discovering and sharing artificial intelligence applications that improve diagnostics, achieve provider well-being, automate tasks and enhance the EMR experience, and his talk “Gen AI in Clinical Practice: Promise, But Much to Improve,” will cover how physicians can leverage AI to relieve burnout, especially with the electronic medical records.
“There is great interest among doctors about how AI might affect them,” Moon says. “There’s great interest among healthcare leaders about making purchases of this expensive software, but you want to get the right one. And there’s a lot of promise by vendors that what they have is the next big thing. But they’ve been promising home runs, and all we’ve seen are singles and doubles at best.”
Moon says that for now, with AI there is greater potential than there are results, but he’s cautiously optimistic. “Artificial intelligence is in its worst state right now,” he says. “It’s only going to get better, and two, three, maybe five years from now it will be something quite impressive.”
Loading ...