As artificial intelligence (AI) continues making inroads in healthcare, radiology stands out as one of the first disciplines where AI has shown promise.

Defined as computer systems able to perform tasks that usually require human intelligence, AI can enhance and expedite human decision-making through pattern recognition, segmentation, quantification, and large-volume data handling. Machine learning (ML), a field of AI, uses statistical techniques to give computer systems the ability to learn from data, without being explicitly programmed. Radiology provides perfect opportunities to apply ML because, thanks to digitization, radiological studies offer ML systems enormous volumes of image data from which to build models.

“You can use AI to seamlessly support existing workflows in order to identify improvements at the point of care,” said Amit Phadnis, chief digital officer at GE Healthcare.

But could these high expectations be dashed, as with the commercial AI craze in the late 1980s? Not this time, say experts.

“Previous collapses of over-hyped AI technology have been called ‘AI winter,’ wrote Curtis P. Langlotz, MD, PhD, professor of radiology and biomedical informatics and director of the Center for Artificial Intelligence in Medicine and Imaging in the Department of Radiology at Stanford University in the September issue of RSNA News. “While today’s enthusiasm for AI can be excessive, these new algorithms have indisputable value to radiology practices, making a severe AI winter unlikely.”

Now, after years of ongoing academic research, commercial clinical systems are appearing from a slew of companies, including big players in AI and medical imaging like IBM and GE Healthcare. No wonder. Medical imaging in AI diagnostics is expected to represent a $19 billion market opportunity by 2025, according to market intelligence company Tractica.¹

A glance at the agenda of the RSNA Annual Meeting reveals how AI and ML are in the spotlight. Both are the focus of plenary sessions, educational courses and special interest sessions.

Impact on professionals

“I see tremendous potential for AI and for machine learning systems to help take away some of the drudgery, some of the rote things that we do every day,” said Charles Kahn, MD, professor of radiology at the Hospital of the University of Pennsylvania, and vice chairman, department of radiology at the University of Pennsylvania.²

Early concerns about AI’s impact on radiology as a profession — “Machine learning will displace much of the work of radiologists,” suggested bioethicist and Affordable Care Act architect Ezekiel J. Emanuel, MD, PhD – increasingly feel misplaced, with conferences like RSNA brimming with AI- and ML-focused content.

“It’s very hard to replace radiologists in the sense that radiologists don’t just do one thing. Most AI algorithms are very specifically tuned for doing one thing,” said Gopal Avinash, PhD, senior data scientist, at GE Healthcare.

At last year’s RSNA, there were more than 1,000 sessions at the Deep Learning Classroom, hosted by NVIDIA’s Deep Learning Institute. RSNA 2017 also debuted a dedicated Machine Learning Showcase, highlighting the latest advances in machine learning and artificial intelligence. At this year’s RSNA, the Pavilion has more than doubled in size, with more than 70 companies participating. Beyond its annual meeting, RSNA said it will introduce several new AI educational programs, including AI-focused Spotlight Courses and a series of AI webinars for radiologists. RSNA will also continue to co-sponsor the National Imaging Informatics Course for radiology residents, and early next year will begin publishing a new journal, Radiology: Artificial Intelligence.

“Change is imminent, and it will bring both challenges and opportunities,” wrote Hricak Hedvig, MD, PhD, in a Jan. 28 article published on Radiology.com. “If we embrace computer science innovations, there are good reasons to believe that technologic advances will increase rather than reduce the importance of our profession.” Or as Langlotz said during a how-to session at the 2017 RSNA annual meeting: “Radiologists who use AI will replace radiologists who don’t.”³

Some even say AI will save the profession, which has seen rising levels of burnout. One of the biggest problems facing physicians and clinicians in general is the overload of patient data to review. Radiology is now ranked as the seventh highest specialty for burnout in the 2018 Medscape National Physician Burnout and Depression Report, compared to 20th in 2017 and 10th in 2016.

Challenges

To be sure, there are challenges ahead. For instance, AI imaging research would benefit from image-sharing networks to feed their ML systems, as well as a standardized use-case language. Likewise, as radiologic databases are expanded, open access to this data and to predictive algorithms will provide a means for comparing competing algorithms. Another important issue is how AI will be integrated into clinical practice and hospital workflows. Tantalizingly, AI-enabled imaging suggests ways for the radiologist to consider not only images, but everything else about the patient, which could blur the boundaries between specialties, such as pathology and imaging.

“When you have AI running transparently and seamlessly on intelligent medical devices at the point of care, you can transform the experience, quality and efficiency of that care,” said Keith Bigelow, senior vice president, Edison Portfolio Strategy, at GE Healthcare.

Like X-ray, 3D CT and MR before it, AI is an innovation that will drive radiology’s value and improve patient care.

References

1. https://www.tractica.com/research/artificial-intelligence-for-healthcare-applications/
2. Quoted at the 2018 ARRS Symposium on AI in Healthcare.
3. https://www.healthimaging.com/topics/imaging-informatics/rsna-2017-rads-who-use-ai-will-replace-rads-who-dont