Hope vs. Fear

The rapidly accelerating wave of adoption of AI-enabled cloud computing has raised fears of job destruction, loss of human control of progressively intelligent systems, environmental damage, loss of privacy, reduction in human connection, diminished human creativity, and amplification of bias, as well as other unanticipated consequences.  That’s a lot to fear, and not without reason.

On the other hand, this wave of technology has raised hopes of accelerated scientific breakthroughs, new drug discovery, personalized healthcare, and democratization of education, as well as reduction of laborious, error-prone human tasks so people can spend more time performing creative, enjoyable, uniquely human activities.  That’s a lot to hope for, and not without reason.

Radiologists have long feared that AI would replace them, a fear exacerbated by the public statements of credible individuals.  Geoffrey Hinton, known as the "Godfather of AI," famously declared in 2016 that machines would outperform humans at reading medical images, adding that we "should stop training radiologists now." Dr. Mitchell Katz, the CEO of NYC Health and Hospitals made headlines by suggesting his system could potentially replace human radiologists entirely with AI technology.  During the latter years of his presidency, Barrack Obama pointed to radiology as an occupation that AI is highly suited for due to its reliance on pattern recognition and X-ray analysis. As Mark Twain famously said in 1897, after a New York newspaper mistakenly published his obituary, “The report of my death was an exaggeration.”  Ironically, I used Chat GPT to help me find the exact statement, which is often misquoted.

Let’s focus on what radiologists can hope AI will do to enhance our profession without destroying our livelihoods.

1. Reporting: AI can help produce, correct, and edit radiology reports, improve the accuracy of speech recognition, consolidate relevant clinical histories and lab data, summarize findings from prior reports, help create report impressions and recommendations, identify finding that are associated with clinical syndromes (and suggest those syndromes), flag measurements outside of the normal range, and answer radiologists questions.  Perhaps AI can put an end to the notorious recommendation, “Clinical correlation is needed,” often heard by referrers as “You figure it out.”

2. Imaging: AI can help scientifically reduce the number of images a radiologist must review, which could help radiologists cope with the explosion of images per exam,  a cause of burn-out.  AI could detect findings, so that radiologists can focus on diagnostic reasoning rather than visual inspection of images. AI can help supervise technologists so that exam quality and the quality of technologist’s notes are improved. AI can assist in reporting statistical probabilities for items in the differential diagnosis, making our reports more useful to referrers. 

More than a decade after experts predicted AI would end the careers of radiologists, here we are, busier than ever and in short supply. Focusing on our hopes for AI-enabled, cloud computing can help us forge a future where technology helps us perform our jobs better and more enjoyably. 

Murray A. Reicher, MD, FACR

CEO Synthesis Health