RSNAI Monthly—March 2026 

The Radiological Society of North America (RSNA) leads the charge in advancing artificial intelligence (AI) as a tool to drive excellence in patient care and health care delivery. RSNAI Monthly is a collection of media briefs to help news outlets stay abreast of RSNA’s latest AI initiatives.

In its March issue, RSNA’s flagship journal Radiology explored how test sets influence the performance of AI for detecting pediatric fractures, the future implications of integrating AI into ablation therapy for liver masses, and how structural MRI-based machine learning can identify biological subtypes of Friedreich ataxia, a progressive neurodegenerative disease. A multinational study found that “deepfake” X-rays aren’t easily distinguishable from the real thing by radiologists—or by AI itself—and an accompanying editorial cautioned radiologists that while the potential to create fabricated images isn’t new, it has lately become more accessible with the widespread availability of generative AI. Meanwhile, a letter to the editor illuminated how evaluating and implementing AI for triage is more complicated than it seems. 

RSNA’s journal Radiology Advances—dedicated to timely dissemination of original research that’s most likely to impact the practice of radiology and patient outcomes—published March articles on AI-enabled opportunistic screening for cardiomegaly on chest CT, deep reinforcement learning agents for identifying CT landmarks in Stanford type B aortic dissection, and a survey that showed large language models can help patients better understand, and feel less anxious about, their lung cancer screening reports.

Studies in RSNA’s journal Radiology: Cardiothoracic Imaging found that deep learning–reconstructed MRI could improve image quality in patients with suspected arrythmias and that a deep learning–based neural network enabled virtual monoenergetic CT imaging and improved the assessment of heavily calcified coronary plaques while reducing blooming artifacts.

Deep learning models for MRI also take center stage in RSNA’s journal Radiology: Imaging Cancer, with applications showing promise for evaluating lymph node metastasis in early-stage cervical cancer and improving diagnosis of hepatocellular carcinoma. 

A special report in RSNA’s journal Radiology: Artificial Intelligence tackled the lack of a single, consistent definition of “machine hallucination” in generative AI, emphasized the need to specify the intended meaning of machine hallucination in radiology, and presented a proposed taxonomy that clarifies when AI errors arise from unsound reasoning versus irrelevant or misleading inputs. 

With its suite of six premier journals, RSNA covers the breadth of groundbreaking research, career-advancing education, and specialty-focused techniques and trends for medical imaging professionals.

The Medical Imaging and Data Resource Center (MIDRC), funded by the National Institute of Biomedical Imaging and hosted at the University of Chicago, is co-led by RSNA, the American College of Radiology, and the American Association of Physicists in Medicine. Held on the third Tuesday of the month, the Seminar Series is an opportunity for members of the medical community at large to hear directly from the MIDRC team. The next session, planned for Tuesday, April 21, will feature research presentations from MIDRC investigators on new and noteworthy advances, and will include a live Q&A session for all attendees.  Free registration is required.

The RSNA Medical Imaging Resource for AI (MIRA) data repository provides annotated medical imaging datasets to support AI research. Developed through RSNA’s ongoing AI Challenge competitions, MIRA datasets help advance innovation in medical imaging and are available at no cost for non-commercial research.
Explore available datasets now.

RSNA’s Annotated Library of AI Systems (ATLAS) now houses more than 230 model cards and dataset cards in 31 subspecialties, giving research communities the tools to make AI research more transparent, understandable and trustworthy. Researchers and developers can publish data cards that describe the key attributes of their AI models and datasets. Users can use ATLAS to discover, evaluate and compare these resources.