RSNA 2023 Day One Highlights: Artificial Intelligence (AI), Radiomics, Regulatory Issues and Health Equity in Imaging

November 27, 2023 — The collective anticipation of the radiology community was palpable from early morning through evening during the first day of the 109th Scientific Sessions and Annual Meeting of the Radiological Society of North America, RSNA 2023. As always, day one did not disappoint in delivering blockbuster news from major industry players, offering the leading experts on hot topics, presenting hundreds of educational and scientific sessions, and closing with a captivating Opening Plenary Session and President’s Address (which will be featured by ITN in a separate session summary with highlights from Matthew Mauro, MD).

Here, then, a summary of lessons learned along the way on this first day, Sunday at RSNA.

AI in Breast Imaging

Setting the stage for the immense amount of discussion and insights on artificial intelligence — and its growing impact on day-to-day radiology, one of the first sessions of RSNA 2023 brought leading minds to a crowded room of participants in the Sunday morning “AI in Breast Imaging” session.

The stellar panel was led by Connie Lehman, MD, PhD, FACR, FSBI, Massachusetts General Hospital, Professor of Radiology at Harvard Medical School; Fredrik Strand, MSc, MD, PhD, Karolinska Institute, Stockholm, Sweden; and Etta D. Pisano, MD, FACR, American College of Radiology (ACR) Division of Cancer Prevention/Chief Research Officer, widely recognized as a pioneer in breast imaging research. ITN will be covering a more in-depth overview of this insightful session in a future feature.

The agenda focused on AI and Mammography Interpretation, with each speaker adeptly covering key areas of interest. Lehman’s Clinical Implementation segment delved into challenges and opportunities, offering valuable historical context on the development of artificial intelligence (AI). Notably, she shined a spotlight on leaders in the field of radiology, including Dr. Fei-Fei Li, author of “The Worlds I See,” describing her as an incredible woman, and mentioned others, such as Geoffrey Hinton, MD, famed for saying AI would take radiologist’s jobs, among others who she described as “brilliant people who saw a different world.”

Strand’s comprehensive but concise scientific research update segment summarized outcomes in the most recent screening trials from the just the past several months. “All of these studies show great promise,” he offered as he shared an overview of the following: MASAI Triage Randomized Trial, published in Lancet Oncology, coordinated by ScreenPoint Medical; the ScreenTrustCAD: AI standalone paired-reader study by Lunit, published in Lancet Digital Health, for which he was co-author along with Dembrower, et al;  and the Clinical Implementation Safety Net study by Ng et al, through Kheiron, recently published in Nature Medicine.

Additionally, regulatory issues and a forward-looking analysis on the pathway to U.S. Food and Drug Administration (FDA) approval and post market testing was offered by Pisano prior to a panel discussion and engaging audience Q&A session. In “U.S. Regulatory Pathway for Breast AI: Current State and Future Directions,” Pisano (expressing her own views, not those of ACR) provided a thorough lesson in current regulation of AI/Machine Learning (ML) software intended uses, limitations on standalone performance testing, reader studies and clinical trials. She noted two final points which were that the EU and UK are ahead of the US in implementing testing AI for clinical practice. She also noted, “The UK National Health Service is going to launch a Real World Evidence AI trial in the next year that will likely be using a design very similar to the one just presented to test the use of Autonomous AI as a second reader in their national breast cancer screening program.”

When an early year resident approached to ask the impressive panel for advice on where to focus his interest, Lehman promptly offered an enthusiastic reply, drawing applause from the audience, when she replied: “Welcome to an incredible subspecialty in radiology…your impact will be extraordinary!”

Imaging AI in Practice Demo

Moving along the AI pathway, another instructive presentation was offered by Alysha Dhami, MD, Stanford Medicine Radiology Resident during an “Imaging AI in Practice” program in the AI Showcase section of the Technical Exhibit Hall. The benefits shown in the demo included how AI models can improve workflow for radiologists, from interpretation to report dictation. Also, it demonstrated how AI models can help radiologists interact with patients and ensure appropriate follow up.

In walking through a typical patient encounter and how “Frank” would experience a range of services from a range of AI-related vendors, Dhami presented an educational demonstration of the ways in which workflow is optimized in ways that are useful for anyone from a first-year resident to an emergency room physician.

“AI models can improve our workflow, and efficiency is key as volumes are skyrocketing,” Dhami noted. Products included in the presentation included AbbaDox, FoviaAI, koios, Laurel Bridge, Milvue, Quera, Siemens Healthineers and Visage Imaging, each offering unique solutions to radiology departments struggling to offer exceptional patient care with often overwhelming workloads.

Takeaways included the following noteworthy recommendations and insights:

The architecture of the overall system for a hospital is critical. Requiring international standards such as IHE, DICOM, HL7, FHIR will enable interoperability. Requiring structured, coded semantic systems such as RadLex, LOINC, and SNOMED will make the AI and other data flow through the system. Lastly, she noted, it is important to ask for all of these in every RFP for every system.

Overcoming Overtreatment for Breast Cancer Patients

In the educational session, “Reducing Overtreatment in Early-stage Breast Cancer,” the panel of presenters covered insights from population data, presented by Nisha Sharma, MBChB, FRCR, who is a consultant breast radiologist at the UK-based Leeds Teaching Hospital NHS Trust, while Bethany Lynn Niell, MD, PhD, Moffitt Cancer Center’s Breast Imaging Section Chief, and NIH/NCI Primary Investigator, shared valuable insight into the promise of radiomics. Moderator Lars J. Grimm, MD, MHS, Associate Professor of Radiology at Duke University School of Medicine and member of Duke Cancer Institute, offered insights from radiology and a range of recommendations on what radiologists can do to reduce overtreatment.

In shedding light on recent data and organizational campaigns to avoid both over diagnosis and over treatment, the panelists noted increasing literature showing decreasing low value care improves patient’s outcomes, as purported by the American Society of Breast Surgeons through its “Choosing Wisely Campaign” and “Radiology in the Era of Value-based Heath Care,” by Brady et al published in Journal Radiology in 2021. The panel described overtreatment and the implications for patient care. They addressed the limitations of current strategies to identify patients at risk for overtreatment, and identified promising new approaches in development that can triage patients away from overtreatment for better long-term outcomes.

“Radiologists create first impressions, and there is a lot we can to do set up the initial framework of care beyond the diagnosis,” noted Grimm. Speaking to that initial point of contact being critical, he added, “We want to provide resources where we can and match the patient’s needs,” as he noted this often means identifying the appropriate surgeon or oncologist to best adapt to the individual patient, and stressing the importance of being a team player working along with other subspecialties.

Niell discussed how the imaging community might use radiomics to address overtreatment in early-stage breast cancers. In defining radiomics, Niell offered this, as she referenced a landmark paper on the topic, aptly titled, “Images are more than pictures, they are data.” Radiomics is defined as “the conversion of images into higher dimensional data that we can then mine for improved decision support…The goal for radiomics is to use standard of care images. We want to use the entire tumor when we can, but we do not need to restrict our radiomics or our quantitative image analysis to the tumor.” She continued, “Another interesting component is radiomics really gives us the opportunity to think longitudinally. Instead of looking at radiomics features on a single image, taken at a single point in time as a snapshot of what that tumor looks like, we have the ability to look at those radiomics features after a series of image studies, and potentially using different types of imaging modalities.”

Health Equity

Stopping by the Radiology Health Equity Coalition booth proved fruitful in gaining a better understanding of its mission, as shared by RHEC Team Lead Carla Brathwaite, with the following comment: “We were founded in 2021, consisting of over 40 major radiology organizations and community health partners across the United States, in an effort to look at healthcare disparities in medical imaging. We have formed valuable partnerships with organizations to look at disparities in screening, as well as in treatment through radiation oncology. Our hope is to really connect with others and look for partnerships to continue to advance health equity in radiology.”

The Imaging Technology News/ITN team also engaged in a large number of visits and interviews with executives and technology leaders across the RSNA 2023 Technical Exhibit Hall, and will report on breaking news, emerging trends, featured solutions and additional education and scientific sessions throughout the meeting.

Follow comprehensive coverage of ITN's RSNA 2023 news by the editorial team here.

More information: www.rsna.org