Radiologist examines AI's potential to transform medical imaging
The emergence of artificial intelligence (AI) in clinical practice is on the horizon. This will be evident at RSNA 2022 in Chicago, which will house numerous AI educational sessions, product demonstrations and an entire wing dedicated to the innovate technology.
While these exhibits are sure to impress, will they answer any of the lingering questions relative to the future of AI? How, exactly, will AI impact radiology practices in the immediate future?
Linda Moy, MD, who is set to become the first female editor of the journal Radiology in January 2023, addressed those questions and more when she spoke to Radiology Business ahead of RSNA 2022 in Chicago. Moy is a professor of radiology at the NYU Grossman School of Medicine and an internationally recognized leader in breast imaging, MRI and AI.
Moy will participate in several presentations during RSNA 2022, including:
- "Understanding and Communicating Artificial Intelligence: Reading, Writing and Reviewing" on Tuesday, Nov. 29.
- "Meet Drs. David A. Bluemke and Linda Moy, Editors Radiology," on Wednesday, Nov. 30.
- "Back to Basics: What Do Rads Need To Know About Radiology AI In 2022?," on Wednesday, Nov. 30.
Speaking with Radiology Business, Moy suggested that the use of AI can help address a multitude of issues today’s radiologists face, but that obstacles remain before research can be translated into real-world practice.
The many potential benefits of AI
One issue AI may be able to help address is radiology's current staffing woes. AI technologies could help with everything from overcrowded emergency rooms to mounting radiologist workloads, she said.
“It can help us better distribute our staff based on hours of peak patient flow, e.g. adding more staff when the ER is busy and fewer staff when it is quiet," Moy explained. "For larger networks of hospitals and practices, it can help mitigate staffing issues across the enterprise."
When it comes to the growing workloads of radiologists—something Moy describes as a big issue— she said that AI could address the problem when used as a decision support tool, in a triage capacity and in assisting with faster acquisition and post-processing of images. She did, however, note that more research needs to be done before this becomes a reality.
Moy also noted that AI can play an important role throughout an image's life cycle. AI could be especially beneficial in protocoling exams, for instance, and in both image acquisition and reconstruction.
“These upstream costs account for most of the cost of an imaging exam," she said.
Moy further elaborated on how AI can enhance practices and workflows in an administrative capacity.
“There are many opportunities for AI to increase our efficiencies beyond image classification and detection of abnormal findings,” she said.
The scheduling of exams, prioritization of exams with critical findings to be read first, worklist management and communication are just some of the other areas where Moy said AI could make a significant impact.
Will radiologists be replaced? Not so fast ...
Recent reports have indicated that misconceptions about the future of AI are deterring many medical students from pursuing careers in radiology, as they believe the technology would render the specialty obsolete. Moy indicated that the use of AI should not impose upon career opportunities, but that it will instead enhance these prospects.
“AI can assist by making healthcare in general more efficient,” Moy said. “I think the fear that AI may replace radiologists turned out to be a false alarm. AI has accelerated novel technology in medical imaging–making it a more attractive field.”
One major issue with many FDA-approved algorithms
Even though a long list of advanced AI algorithms have been approved by the FDA, Moy highlighted the fact that there may be a lot of work to be done before they can be trusted with large, diverse patient populations. Many of the current FDA approvals, she said, have been based on small retrospective studies. This makes them less generalizable to larger, more diverse populations, which could prove as a roadblock in bringing AI into routine practice.
“The biggest obstacle is to see if these AI tools work in a variety of clinical settings and heterogeneous patient populations," she said.
Moy is set to present in numerous sessions at this year’s annual RSNA meeting.
Additional coverage of RSNA 2022 is available here and here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She joined Innovate Healthcare in 2021 and has since put her unique expertise to use in her editorial role with Health Imaging.