In a very stimulating conference, Eric Topol, MDBestselling author and practicing cardiologist at the Scripps Clinic in San Diego and editor-in-chief of medical landscape, told a standing-room-only audience at the Dec. 2 plenary session in RSNA24: this year’s conference of the Radiological Society of North America—That artificial intelligence will transform the practice of medicine in the coming years.
Speaking to a standing-room-only audience at the Arie Crown Theater in Chicago’s vast McCormick Place Convention Center, Dr. Topol, author of the 2019 bestseller deep medicine, He took his audience of radiologists and others involved in radiology through the evolution to date of artificial intelligence, and then predicted, based on the progress so far, what would happen next.
Topol began by contextualizing the current moment, noting that 800,000 Americans die or are severely disabled each year due to misdiagnoses; What is on the horizon, he emphasized, is a new era in which artificial intelligence tools will help doctors better diagnose and treat, and even predict the onset of diseases. He said the fundamental work done over the past few years in developing algorithms and working with large language models has laid the foundation for massive change. For example, data collected from huge amounts of data and images is already leading to better diagnoses, such as in gastroenterology, where gastroenterologists are already using AI-enabled endoscopy to detect more polyps than they expected. they could detect previously. And data is being collected even from diagnostic images such as X-rays, creating huge data lakes that are used to support medical diagnostic processes. He called this phenomenon “Machine Eyes,” the collection of data that, when analyzed and used to support clinical decisions, can improve diagnosis. Surprisingly now, studies are finding that analysis based on chest x-rays can lead to the diagnosis of a surprising variety of diseases, including diabetes. He cited a September 2023 study based on the analysis of 1.6 million retinal images collected in the United Kingdom that produced groundbreaking predictive diagnoses.
Now, Topol told his audience, medicine is on the verge of being able to use two types of multimodal AI: one based on text, voice and images, and the other based on human data. “Where can multimodal AI take us?” asked. “In the future, a very different level of precision and accuracy can be achieved in medicine,” he predicted. “For example, hospital at home may be more of a consideration in the future,” as the analytics needed to support these cutting-edge forms of care delivery will become increasingly available.
What’s more, Topol reported that “four basic models in pathology were published last year” in clinical journals. They will allow you to achieve a “diagnosis based on a complete image”.
In the meantime, he said, what is becoming clear is that “AI does a really good job with your text in terms of completeness, correctness and conciseness. “AI reports are more accurate, easier to understand and more complete than reports produced by doctors.” It also noted a couple of studies that have concluded not only that AI does a better job of diagnosing than human doctors, but two studies have found that AI itself actually does a better job of diagnosing than AI. + humans. That result, however, he quickly added, is probably related to the fact that the studies were “artificial” artificial tests, not based on real patient care situations. It is interesting to note, however, that AI appears to promote the expression of empathy among doctors.
Ambient intelligence and a new range of capabilities
Topol noted that “generative AI, not just NLP [natural language processing]Audio notes can be converted into various patient notes. In fact, it’s more accurate than regular EHR notes. You can schedule follow-up appointments, order prescriptions, etc. And it can even train doctors to be more empathetic and better communicators.”
Above all, Topol said, AI can help give doctors “the gift of time,” through “keyboard liberation,” the ability to synthesize patient data, the ability to preview all images in the primary examination and automated diagnosis. of routine conditions.
And one of the biggest types of potential, Topol said, is longitudinal data that can facilitate “individualized medicine from the preuterus to the grave.” It’s that kind of data, which was included in recent research at the Weizmann Institute in Israel, that is uncovering the “biological clocks” within human bodies that age at different rates. These same data could help individualize the diagnosis in oncology; For example, he noted, pancreatic cancer is one of the most difficult cancers to detect at a very early stage of its progression; Data analysis could suggest which patients might be at higher risk. And research continues on the use of plasma proteins, obtained with just “a couple of milliliters of blood,” which can detect the risk of disease.
The main obstacles right now to advancing this whole area around AI, Topol said, are the following: the medical community’s resistance to change; refund issues; regulatory challenges; the need for greater transparency; the need for convincing evidence; build trust between doctors and the public; and implementation challenges.
