Choosing a Specialty in the Age of Artificial Intelligence

Marlene Busko


September 13, 2018

When selecting a career path, should medical students worry that in specialties where image interpretation is key they might soon be replaced by a machine with artificial intelligence (AI)? After all, in two recent preliminary studies (Esteva et al, Nature 2017 and Haenssle et al, Ann Onc 2018), a "trained" computer algorithm was as good as or better than expert dermatologists in distinguishing between pictures of harmful versus harmless skin lesions.

Are these developments scaring students away from these specialties, or should aspiring radiologists and dermatologists look forward to a time when algorithms, machine learning, and AI free them from mundane aspects of clinical practice and open up new ways to deliver better patient care? Medscape interviewed a few dermatologists, radiologists, and residents to get their thoughts.

Stop Training Radiologists Now?

Eliot Siegel, MD, professor and vice chair of research information systems in the Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, and chief of imaging services at the VA Maryland Health Care System, in Baltimore, told Medscape that some experts predicted that machines would replace radiologists within 3 years.

Professor Andrew Ng, from Stanford University, in California, famously said in The Economist, "A highly trained and specialised radiologist may now be in greater danger of being replaced by a machine than his own executive assistant."

And Professor Geoffrey Hinton, a pioneer in artificial neural networks who does research for Google and the University of Toronto, tweeted 2 years ago, "We should stop training radiologists right now, in 5 years #deeplearning will have better performance."

Siegel vehemently disagrees. He emphasized to Medscape, "I feel very strongly that radiologists should have no concerns about being replaced by a computer. At least for the next 25 years of their medical careers.... Computers will be helping [clinicians] to make better quantitative diagnoses with things that are repetitive and relatively low-level," thus allowing "them to be freed up to make judgments about cases and better correlate clinical information."

Residents need not worry that they will be replaced by a machine any time soon, Graeme M. Lipper, MD, a dermatologist at Danbury Hospital, Connecticut, echoed that sentiment to Medscape. "Most young radiology residents want to do interventional [radiology] anyway," he pointed out, and "someone would still have to explain the diagnosis and treatment options to the patient and perform the surgery."


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.