By Aine Cryts

His goal may seem simple and straight-forward now, but Eliot Siegel, MD, FSIIM, hoped back in the early 1990s to make radiology images available anytime, anywhere to everyone who was authorized to see them. It wasn’t so simple, at all. Siegel also wanted to harness the potential of computer workstations to change the entire practice of diagnostic imaging.

How was he going to do that? By setting up the first filmless radiology department, Siegel, professor of radiology at University of Maryland School of Medicine and chief of imaging services at VA Maryland Health Care System, tells AXIS Imaging News.

“The emergence of digital, rather than film-based image interpretation, meant that more complex imaging studies could be reviewed, which made multi-slice and channel CT practical as well as MRI studies that contained increasingly large and complex datasets,” he adds.

AXIS recently asked Siegel about his excitement about technology, including artificial intelligence (AI), and its ability to solve problems in healthcare. Here’s what he had to say.

AXIS Imaging News: Setting up the first filmless radiology department sounds pretty ambitious. What else was involved?

Siegel: The other initial goal of the transition to digital systems was to take advantage of advanced visualization, image analysis, computer-based detection, and diagnosis and decision support. While advanced visualization has made major strides over the past 30 years, the implementation of AI systems has lagged far behind, unfortunately.

I’m currently excited about the tremendous opportunities we finally have to turn our vast repositories of digital imaging data into algorithms for detection, diagnosis, and treatment for patients. That’s in addition to the major potential for improved efficiency, productivity, and stress reduction as well as major enhancements in image quality. Also exciting to me is the potential to reduce both radiation dose and imaging time.

AXIS: As a trained and practicing radiologist, what special skills do you bring to this work?

Siegel: There are many reasons that a radiologist or at least someone with extensive clinical training in radiology can have a major impact on advancement of the technology. 

AI systems will provide the greatest value when they combine machine learning with clinical decision-making. AI algorithms that detect findings, such as lung nodules or intracranial hemorrhage or bone age, are increasingly becoming a commodity. The differences between developers’ results are rapidly shrinking.

Clinical decision-support suites or packages that provide lung screening software or stroke- or trauma-based detection with recommendations will increasingly become preferred for diagnostic imaging. These packages will require heavy input from data scientists, software developers, and clinicians to determine the greatest need. In addition, data scientists, software developers, and clinicians will need to isolate the greatest pain points and the right solutions that can provide the greatest added value. Only a radiologist or another clinician knows their own clinical workflow in detail and can articulate this to refine or even reinvent these clinical processes. 

AXIS: You’re pretty enthusiastic about efforts to use AI in healthcare. What’s your advice for other radiologists who want to learn about this technology?

Siegel: Get first-hand experience with AI systems. AI platforms are emerging from vendors that provide advanced visualization or speech recognition or even image registration; that’s in addition to offerings from more traditional PACS and modality vendors. Many vendors allow radiologists to try out the AI software on a trial basis using their own data in their own facility. The more experience radiologists have with the limitations and features of these systems, the more they’ll be able to contribute ideas and engage in collaborations or even create their own startups.

AXIS: What are some other resources for radiologists to learn about AI?

Siegel: The American College of Radiology’s Data Science Institute is a great source of information. It offers numerous pathways to get involved, including tools to help mine imaging databases. Also, the Society of Imaging Informatics in Medicine offers extensive sources of information, including the annual meeting and its Conference on Machine Intelligence in Medical Imaging. Online free resources, such as Andrew Ng’s Coursera lectures in machine learning and deep learning, are abundant and excellent. 

Aine Cryts is a contributing writer for AXIS Imaging News.