Journal highlights

Diagnostic reference levels (DRLs) and achievable doses (ADs) help facilities evaluate radiation dose levels and optimize CT imaging protocols. However, national benchmarks are updated infrequently and may not reflect advances in CT technology, machine learning based reconstruction methods and dose monitoring.

In a new retrospective observational study published in Radiology, Kalpan M. Kanal, PhD, University of Washington School of Medicine in Seattle, and colleagues analyzed approximately 5.2 million adult CT acquisitions from 592 facilities across 10 examination categories, nearly four times more than prior benchmark studies.

Compared with 2014 benchmarks, national CT volume dose index ADs decreased by 9.3%, while DRLs declined by 21.8%. Dose reductions were greatest for chest CT examinations with and without contrast material, at 31.2% and 26.7%, respectively, and smallest for head CT without contrast material at 3.5%.

“The scale, consistency and automation of this framework support population- level benchmarking for quality improvement and protocol optimization,” the authors conclude.

Read the full article, “A Decade of Change in Adult CT Radiation Doses: 2025 U.S. Diagnostic Reference Levels,” and the accompanying editorial, "Good News about CT Doses."

Follow the Radiology editor on X @RadiologyEditor.

Navigating PI-RADS 3 Lesions: Reducing Uncertainty and Guiding Prostate Cancer Diagnosis Decision Making

Prostate Imaging Reporting and Data System (PI-RADS) category 3 lesions remain among the most challenging findings in prostate MRI, with an equivocal likelihood of clinically significant prostate cancer and uncertainty about whether biopsy is warranted. Under PI-RADS v2.1, assessment relies on multiparametric MRI, including T2-weighted, diffusion weighted and dynamic contrast-enhanced sequences, with scoring criteria that vary by prostate zone.

In a review published in RadioGraphics, Refky Nicola, DO, MSc, SUNY Upstate Medical University in Syracuse, NY, Eduardo Thadeu de Oliveira Correia, MD, PhD, University Hospitals Cleveland Medical Center, and colleagues examine limitations, factors and strategies shaping PI-RADS 3 lesion management. The authors cite contributors to PI-RADS 3 assessments, including suboptimal image quality, postprocedural or treatment-related changes, noncontrast imaging and reader experience.

They also emphasize that ancillary biomarkers can support risk-based, MRI-directed pathways that help avoid unnecessary biopsies while reducing missed clinically significant cancers.

“A thorough understanding of the limitations related to, factors contributing to, and evolving management strategies for PI-RADS 3 lesions enables physicians to make better informed clinical decisions, reduce unnecessary interventions, and improve the detection of clinically significant prostate cancer,” the authors write.

Read the full article, “Navigating PI-RADS 3 Lesions: Reducing Uncertainty and Guiding Prostate Cancer Diagnostic Decision Making.” This article is also available for CME on EdCentral. 

Follow the RadioGraphics editor on X @RadG_Editor.

Using AI to Estimate Lung Nodule

The growing adoption of low-dose (LD) CT screening for lung cancer is increasing radiologists’ workload. An AI system may help reduce that burden by improving the assessment of indeterminate pulmonary nodules. The Lung Nodule Analysis 2025 (LUNA25) challenge is a public, international competition designed to benchmark AI algorithms for lung nodule malignancy risk estimation on LDCT scans.

In a study published in Radiology: Artificial Intelligence, Dré Peeters, MSc, Radboud University Medical Center, Nijmegen, the Netherlands, and colleagues compared the best-performing AI system from the LUNA25 challenge with 65 radiologists using a subset of 300 indeterminate nodules from three European lung cancer screening trials.

The AI system achieved a significantly higher AUC than the radiologists’ average performance (0.78 vs. 0.70) and outperformed 95% of individual radiologists.

At matched specificity, the AI system correctly identified 12% more malignant nodules than radiologists. At matched sensitivity, it generated 20% fewer false-positive results.

“Prospective validation should further investigate the use of this system as a decision-support tool and assess the feasibility of its integration into current workflows,” the authors conclude. 

Read the full article, “Benchmarking of AI and Radiologists for Indeterminate Lung Nodule Malignancy Risk Estimation at Screening CT: The LUNA25," and invited commentary, "What LUNA25 Teaches Us about AI for Lung Cancer Screening."

RadioGraphics Monograph Focuses on Disease Screening

The October RadioGraphics monograph will explore the role of imaging in disease screening. Published since 1999, the annual monograph features focused issues across radiologic subspecialties, offering up-to-date reviews of state-ofthe-art topics for radiologists.

This year’s collection will examine screening across body systems, with articles on cardiovascular disease, pediatric cancer predisposition syndromes, colorectal cancer, breast cancer in pregnant and lactating patients, opportunistic CT screening, gynecologic malignancies, hepatocellular carcinoma, high-risk newborns and infants, lung screening and pancreatic cancer screening.

Learn more and explore past issues of the RadioGraphics monograph.

Become an RSNA Peer Reviewer

Peer Review Week, Sept. 14–18, is a reminder of the essential role reviewers play in advancing the quality and integrity of research. This year’s theme, “Peer Review Capacity: Volume, Speed, and Quality,” underscores the need for thoughtful, timely review as scholarly publishing continues to evolve.

Help advance radiologic science and support your specialty by becoming an RSNA peer reviewer. Opportunities are available to review RSNA annual meeting abstracts and manuscript submission for all six of our journals.

Explore peer review opportunities