5 Things to Watch in 2025
By Dave Yeager
Radiology Today
Vol. 26 No. 1 P. 24
News and Notes from RSNA 2024
There was plenty of hustle and bustle at RSNA 2024. Although attendance has yet to reach prepandemic levels, it increased to nearly 39,000 registrants. The theme of this year’s conference was Building Intelligent Connections, and it resonated from the plenary sessions to the exhibit halls. There was plenty of hustle and bustle at RSNA 2024. Although attendance has yet to reach prepandemic levels, it increased to nearly 39,000 registrants. The theme of this year’s conference was Building Intelligent Connections, and it resonated from the plenary sessions to the exhibit halls.
In the president’s address, RSNA 2024 President Curtis P. Langlotz, MD, PhD, talked about the prevalence of AI in radiology and noted the importance of using appropriate data to prevent data bias. He pointed out that patients are often willing to share their personal data to train AI and that data based on patient models will help practices choose which AI algorithms suit their needs. Langlotz believes that AI isn’t coming for anyone’s job but is an essential part of medical practice that will allow radiologists to delegate repetitive tasks, leaving more time for patient care. As he said, “Radiologists who use AI will replace radiologists who don’t.”
In the opening plenary session, “The Only Way to Predict the Future Is to Create It,” Nina Kottler, MD, MS, expanded on the ways that AI may work in radiology practice. She advocated for AI augmenting radiologists’ performance. The challenge and the opportunity, Kottler said, is that AI and humans have different biases. By mitigating AI and human biases, physicians can benefit from the best that each has to offer. Kottler noted the increasing use of generative AI models—models that synthesize data from existing data (see page 10 for more on generative AI)—and she said the best way to approach AI is to put a team in place that can create a data and AI ecosystem within the health care organization.
In another plenary session, “Social Artificial Intelligence in Hybrid Systems of Humans and Machines,” Nicholas Christakis, MD, PhD, MPH, offered more detail about how humans and AI can and/or should interact. To optimize AI, Christakis said, it’s important to consider how machines influence social interactions. Depending on the type of AI, it can improve or degrade people’s ability to work together. He added that AI should support human reciprocity, not replace it, and the nature of our connections affects the properties of a group.
Just as AI is changing medicine, AI itself is changing. In “AI’s Transformation of Medicine,” Eric Topol, MD, discussed how AI is being used in medicine and where it may be headed. Not only are data volumes growing, but clinicians are able to get more data from existing tests than ever before. Topol expects to see more large language models, generative AI algorithms, and multimodal AI algorithms— algorithms that can process and integrate multiple types of data simultaneously—although he noted that the FDA has not yet approved any multimodal AI algorithms. Topol believes AI can help facilitate personalized medicine, citing a study from Norway that enabled researchers to ascertain patients’ pancreatic cancer risk based on EHR data, but he cautioned that AI transparency—the ability to demonstrate how an algorithm reached a conclusion—is essential for ensuring its usefulness. He also made the point that poorly implemented AI can increase, rather than alleviate, radiologist burnout.
Burnout has been the topic of much discussion over the past few years, and Carrie Cunningham, MD, MPH, talked about its toll in “Removing the Mask, v. 2.0.” Cunningham’s message is that burnout and mental health concerns are significant issues for physicians, and help is available. As she noted, self-care is health care, and it is essential.
Another interesting session was the “Fast 5” presentation. It’s actually five short presentations on emerging topics. In keeping with the theme of building connections, there were presentations on professional coaching by Tessa Cook, MD, PhD, who is a member of Radiology Today’s editorial advisory board; opportunistic imaging by Abhinav Suri, BA, MPH; determining professional responsibility when AI fails by Niraj Nirmal Pandey, MBBS; community-engaged research to advance health equity by Lauren Groner, DO; and using social media to educate and advocate by Robyn Gartner Roth, MD.
Many of the people I spoke with in the exhibit hall echoed ideas that were highlighted in the sessions and, as always, there was much to see. Without further ado, here is what I found noteworthy:
1 Workflow Orchestration and Reporting
Workflow orchestration and reporting platforms were in abundance. As previously mentioned, radiologist burnout is a significant concern, and vendors are working to streamline reading and reporting. There were many all-purpose platforms, as well as some that were specialized for breast imaging. AI is incorporated in many ways, including helping to spot notable imaging features. To produce accurate and relevant results, AI requires big data, and there was an abundance of big data energy all over McCormick Place. I asked several people how they address data bias and data drift, and the consensus was basically, “We train our algorithms on A LOT of images, and we’re constantly adding more.”
I found some of the reporting platforms particularly interesting. They are getting progressively better at generating reports from dictation. Several can also include recommendations and identify mismatches automatically, and some vendors are beginning to incorporate generative AI to assist in report creation.
2 Workflow Assistance for Rad Techs
Just as burnout is an issue for radiologists, it is also a significant burden for technologists. Repetitive motion injuries are common, and the demand for imaging is outpacing the number of technologists. Two areas that featured significant ergonomic upgrades were ultrasound and mammography, and at least one vendor has a new mobile C-arm that is notably easy to maneuver.
Ultrasound requires a great deal of movement, and consoles tend to have quite a few buttons. Many newer consoles have reduced the number of buttons and streamlined the preset functions to make them faster and easier to use. Additionally, several new models make it easier for less experienced technologists to obtain images.
Some new or redesigned mammography systems also reduced the number of buttons and made it easier to select protocols. There have been advances in breast compression, as well, that reduce the ergonomic strain on technologists, not to mention making the experience easier for patients.
Along with physically helping technologists, some vendors have the capability to allow more experienced technologists to remotely assist less experienced technologists. Additionally, some MR scans can be done entirely remotely, allowing a single technologist to scan patients at multiple facilities.
3 Opportunistic Imaging and Risk Prediction
Opportunistic imaging appears for the second year in a row on this list, and for good reason. In addition to the aforementioned Fast 5 presentation, many people are looking for ways to do more with less. It’s no wonder: Imaging volumes are growing faster than ever so why not use one imaging exam to screen for multiple health conditions? It’s not a new idea, but improving technology is making it a more viable option.
A different but related aspect of our ability to glean more information from imaging exams is risk prediction. This shows up particularly in mammography, but it is also being applied in other areas, such as cardiac imaging. As data models are refined, risk prediction could be a significant marker on the road to personalized medicine.
4 Follow-Up Recommendations
Of course, initial imaging exams are not always conclusive, which necessitates follow-up imaging. Communicating these recommendations has been an ongoing challenge, particularly in underserved areas. One vendor I spoke with estimated that 20% of imaging exams require follow-up. Several vendors are attempting to address this issue. Whether it be reaching out directly to patients or making information available in patient portals, several people I spoke with identified this as a critical need. Expect to see more effort toward closing communication gaps.
5 The Value of Partnerships
The number of partnerships that were announced caught my attention this year. In any year, there are a fair number of them but, this year, it seemed as though there were more hook-ups than Bachelor in Paradise. It makes sense when you consider that technology and medical imaging are becoming increasingly specialized, and medical imaging systems require multiple specialized components. In many cases, it’s more efficient— and cost effective—to partner with someone who specializes in something you need, rather than trying to build from scratch.
That’s all for this edition. I’ll see you at RSNA 2025.
— Dave Yeager is the editor of Radiology Today.