Artificial Intelligence in Radiology

Jan Beger

There is a lot of hype around Artificial Intelligence (AI) in medical imaging recently. In the radiology community, there is concern over what the technology will mean for the future of the industry.

Last year, Toronto University professor and AI pioneer Geoffrey Hinton warned, AI will likely read medical imaging better than radiologists, resulting in unemployment across the industry, and that medical schools should stop training radiologists.1 Earlier this year, Vinod Koshla, a US venture capitalist and co-founder of Sun Microsystems stated in an interview that, “…the role of the radiologist will be obsolete in five years.”2 Unfortunately, this did not take into account the fact that being a radiologist goes way beyond the interpretation of images. AI is unlikely to replace radiologists anytime soon, rather it will increase the value they provide.

Increasingly, AI will be used in reading workflows and the interpretation of radiology images, to address the following challenges:

  • Demand: In most countries, there are not enough radiologists to meet the demand for diagnostic imaging services—radiologists are operating at, or near capacity. The situation will get worse as the demand for diagnostic imaging services grows, populations age and chronic diseases grow at a faster rate than new radiologists enter the domain.3 4 Radiologists have to deal with data overflow as radiology exams become more and more complex and the number of images per study grows. This has led to medical image interpretation becoming a bottleneck in healthcare today. Digital technology has helped to speed the imaging process, but this has not been enough to counter the increased demand and the complexity of imaging studies.
  • Diagnostic accuracy: The average radiologist reads approximately 15,000 cases per year. If you assume an error rate of 4 percent, this means approximately 600 of those cases are misinterpreted.5 Also, 50 percent of the women who get annual mammograms over a ten year period have a false-positive finding. 6

AI is going to tackle these problems and likely will change the way radiology is practiced in the future. Radiology, like other specialties, has to plan strategically for a future in which AI is part of the diagnostic team.

So, how could this technology impact the role of the radiologist in the future? Radiologists and AI must be thought of as complimentary. Radiologists must allow themselves to be displaced by algorithms. Routine, repetitive, time-consuming and resource intensive tasks should be handed over to AI. AI will provide referenced guidelines with similar pathologies and clinical context, and it will prioritize reading work-lists by auto-detecting abnormal cases.

The responsibilities of the radiologist may shift from extracting information from images and other clinical data to managing information extracted by AI in the clinical context of the patient.

AI will make the radiologist faster and more precise, it will take the role of an intelligent clinical assistant, to boost efficiency, productivity and diagnostic accuracy.

One example: Experienced clinicians today can spend 60-90 minutes physically drawing contours on a desktop to calculate blood flow. By contrast, software developed by ArterysTM uses deep learning algorithms to accurately quantify blood flow in 15 seconds, enabling clinicians to focus on higher-order cognitive tasks in caring for the patient.7

While we believe the future of AI in medical imaging is bright, it is far from clear how it will transform radiology in the short term.

I’d love to hear from you. Don’t hesitate to contact me –

3 Radiology in the UK: the case for a new service model – Sep 2014 – ponse_to_Dalton_Review.pdf – The Royal College of Radiologists
4 U.K. steps up its bid to entice overseas radiologists – Mar 2015
– 1021 – Frances Rylands-Monk
5 Bhargavan M, Kaye AH, Forman HP, Sunshine JH. Workload of radiologists in United States in 2006-2007 and trends since 1991-1992. Radiology. 2009;252:458-467.
6 Jena AB, Seabury S, Lakdawalla D, Chandra A. Malpractice risk according to physician specialty. N Engl J Med 2011;365(7):629–636. CrossRef, Medline.
7 Arterys is a trademark of Arterys, Inc.

Leave a Reply

Your email address will not be published. Required fields are marked *