Since the release of large language models such as ChatGPT in November of 2022, medicine has been transformed in ways previously unimaginable by using artificial intelligence (AI) as a tool to solve problems. The use of natural language processing (NLP) in a transformer model allows physicians to explore the massive amount of data in a patient’s electronic medical record to identify key opportunities to improve patient care. In addition, machine learning algorithms applied to images of the retina, chest x-rays, pathologic biopsy specimens, and imaging of the skin identify diseases in a way that is currently unexplainable.
As of October 2024, the FDA has approved over 950 AI medical devices, and as of this year, most hospitals and health systems are beginning to deploy these AI algorithms initially in the form of ambient listening systems, which create a progress note in the patient’s electronic medical record. Over 75 percent of FDA-approved AI as medical devices are in radiology, and at the recent RSNA meeting in Chicago, it has become clear that radiologists worldwide are starting to deploy this technology for patient care. Cardiologists have AI embedded in the systems that they use to perform cardiac imaging, especially for CT scanning of the heart. AI can look at an ECG and determine not only the ejection fraction without an echocardiogram but also the presence of diabetes, thyroid disease, CKD, and anemia. This is possible because AI can detect patterns at a level that we are unable to recognize.
Much of this work was pioneered in the fields of ophthalmology. An AI analysis of an image of the retina can diagnose not only diabetic retinopathy but also recognize Parkinson’s disease and Alzheimer’s disease long before any signs or symptoms are clinically apparent. AI can analyze a pathology image and predict the prognosis of cancer and is currently being used during colonoscopy to detect polyps that would have otherwise been missed. The use of a transformer model such as ChatGPT-4 will give physicians the ability to make subtle diagnoses of rare diseases in ways never before possible.
The 2024 Nobel Prize in both chemistry and physics were recently awarded to scientists for creating the foundation of neural networks (John Hopfield and Geoffrey Hinton) as well as the ability to predict protein folding (David Baker, Demis Hassabis, and John Jumper), which will lead to new drug discovery. As a physician-scientist, this moment in time is truly magical as life-altering breakthroughs are happening at a pace unheard of in the world of medical research.
As a fellow of the American College of Physicians, it became clear that a professional society for physicians working in the AI space is needed. Hence, in August 2024, we launched the American College of Artificial Intelligence and Medicine, which is a not-for-profit professional home for physicians committed to an evidence-based medicine approach to both research and education at the intersection of artificial intelligence and medicine.
Melvin Speisman is an internal medicine physician.
