Radiology Advances Podcast | RSNA

This episode discusses a study from Johns Hopkins University in the United States that introduces a brain age predictor from MRI, designed to stay accurate across different scanners and cohorts. The model estimates a person's age from a brain scan to within about four years, and holds that accuracy on an independent external dataset. The gap between predicted and actual age rose steadily from heal...

This episode covers a study from the Mallinckrodt Institute of Radiology at Washington University in St. Louis evaluating whether patient-specific lung shunt fraction measurement is necessary for every yttrium-90 selective internal radiation therapy case. Across 354 cases, the authors propose a new pretreatment metric called LSFbound — derived from liver mass, lung mass, and dose thresholds — that...

This episode discusses a study from New York University evaluating whether deep learning can predict acute pancreatitis severity from contrast-enhanced CT acquired within 24 hours of admission. Using self-supervised pretraining on about 12,000 unlabeled scans followed by supervised fine-tuning, the model achieved an AUROC near 0.89 for severe pancreatitis on both an internal NYU test set and an ex...

This episode discusses a study from the University of California, San Francisco in the United States that tested whether GPT-4o-generated patient-friendly summaries improve comprehension of lung cancer screening CT reports. In a within-subjects survey of 1,815 adults across Lung-RADS 1, 2S, and 4B vignettes, the summaries significantly improved objective comprehension and reduced anxiety for all t...

This episode explores a study from the University of Texas Southwestern Medical Center and MD Anderson Cancer Center in the United States that clinically validates an FDA-cleared AI tool for measuring total cardiac volume on non-contrast, non-gated chest CT. Across 307 patients with paired echocardiography, the AI discriminated normal from abnormal cardiac volume with an AUC of 0.81 in men and 0.7...

This episode explores a study from the Emory Sports Performance and Research Center and the University of Lausanne that determined how few annotated MRI exams are needed to train a reliable deep learning model for thigh muscle segmentation. Using the nnU-Net framework with incrementally larger training sets, the researchers found that just 20 high-quality annotated subjects produced clinically acc...

This episode discusses a study from UCLA in the United States that used federated learning to train a deep learning model for automatic segmentation and quantification of visceral and subcutaneous abdominal fat in children using free-breathing 3D MRI. By leveraging a larger adult dataset alongside a small pediatric cohort, the model achieved strong agreement with expert manual segmentation in unde...

This episode reviews a proof-of-concept study from Mayo Clinic Minnesota on the use of ferumoxytol-enhanced MRI for detecting gastrointestinal bleeding after a comprehensive conventional workup has been negative. We examine how this blood pool agent's prolonged intravascular half-life addresses the diagnostic challenge of slow and intermittent GI bleeding, and discuss the clinical implications for...

This episode reviews a study from the ROADMAP Group evaluating deep reinforcement learning for automatic aortic landmark localization in Stanford Type B aortic dissection — examining whether AI can match expert human performance for a task critical to treatment planning and long-term surveillance. Deep reinforcement learning for automatic anatomic CT landmark localization in Stanford Type B aortic...

This episode explores a technological advance from Johns Hopkins in the United States that improves diagnostic ultrasound for breast masses. By combining short-lag spatial coherence imaging with an objective metric called generalized contrast-to-noise ratio, the researchers achieved a dramatic boost in diagnostic accuracy—especially in dense breast tissue—while reducing variability among radiologi...

This episode highlights a study from Korea using deep learning to generate synthetic contrast-enhanced brain MRI images—without injecting contrast agents. The model accurately segmented the choroid plexus and matched real contrast-enhanced scans in volume analysis, offering a potentially safer, scalable tool for neuroimaging. Automated synthetic contrast-enhanced MRI improves  choroid plexus segme...

This episode explores a study from Radiology Advances tackling one of AI's toughest challenges in medical imaging: consistent pancreas segmentation across CT scans. The authors benchmarked multiple models against multi-reader human consensus and introduced a new metric, Fractional Threshold (FT), to measure robustness. Their human-in-the-loop workflow flagged just 5% of cases for expert review, ma...

This episode explores a study from Radiology Advances challenging FDA's acoustic output limits for liver ultrasound elastography for obese patients. The authors tested the exam at a mechanical index of 2.5, well above the 1.9 regulatory ceiling, and found no liver injury using stringent biochemical criteria. The payoff: a 29.2% reduction in measurement variability and 40% fewer failed attempts in...

This episode features a cutting-edge study from Radiology Advances exploring Deep Silicon Photon-Counting CT (DS-IPCCT) for liver fat quantification. Using in silico models, the investigational system demonstrated high spectral accuracy, robust material decomposition, and low error rates—potentially overcoming key limitations of conventional CT and MRI.  Liver fat quantification using deep silicon...

This episode explores Radiology Advances research on RadGPT—a hybrid AI system combining image analysis with a language model to interpret knee radiographs. Built on 77,000 images, the system incorporates mandatory human review, dramatically improving diagnostic accuracy and report quality. Host commentary highlights its potential as a diagnostic assistant for trainees and an efficiency tool for e...

This episode covers a study from Radiology Advances evaluating deep learning–accelerated MRI across routine neuroradiology exams. Using Siemens' Deep Resolve, scan times were cut by over 50% without sacrificing diagnostic image quality. Host commentary explores reader preferences, artifacts, and when DL-MRI may be best suited for clinical use. Deep learning MRI halves scan time and preserves image...

This episode discusses a study from Radiology Advances evaluating contrast-enhanced CT as a non-invasive alternative for lung shunt fraction (LSF) estimation in hepatic radioembolization to the current standard, 99mTc-MAA nuclear medicine imaging. The proposed CT-based method showed strong correlation with standard MAA-based LSF, offering a faster, safer, and potentially more accurate planning app...

This episode covers a study in Radiology Advances evaluating deep learning–accelerated T1 MPRAGE MRI in patients with memory loss. The approach cut scan time by more than half while preserving image quality and measurement accuracy—offering faster, more comfortable imaging for dementia care and longitudinal follow-up. Deep-learning-accelerated T1-MPRAGE MRI for  quantification and visual grading o...

This episode spotlights a study from Radiology Advances introducing a fully automated deep learning pipeline for myocardial infarct segmentation on late gadolinium enhancement cardiac MRI. Developed at the Medical University of Innsbruck, the model showed near-perfect agreement with human experts and even outperformed manual segmentations in blinded qualitative review. Deep learning pipeline for f...

A prospective study evaluates ultrasound-derived fat fraction (UDFF) as a tool to monitor hepatic steatosis after bariatric surgery. Host commentary unpacks how UDFF may offer a non-invasive, accessible, and quantitative alternative to MRI-PDFF and liver biopsy, and highlights UDFF's clinical potential for routine liver fat surveillance.   Quantifying changes in steatotic liver disease after baria...

A multi-center study evaluating an AI model for automated CT segmentation of intracerebral hemorrhage, intraventricular hemorrhage, and perihematomal edema. Host commentary highlights how the deep learning tool delivers near-expert accuracy in under 20 seconds—dramatically reducing time and enhancing precision in acute stroke care.   Cross-institutional automated multilabel segmentation for acute...

A prospective randomized trial compares robotic versus manual needle insertion for CT-guided intervention. Host commentary summarizes the results that show the robotic system matched manual accuracy and clinical success rates while significantly reducing radiation exposure to the interventionalist. The discussion touches on clinical implications for workflow, safety, and the evolving role of robot...

MRAnnotator is a deep learning model that segments 44 anatomical structures across diverse MRI sequences. Developed at Mount Sinai, it shows strong generalizability across scanners and sites, outperforming existing models. Host commentary summarizes the model development and datasets and explores its impact on AI development, annotation workflows, and multi-center research. MRAnnotator: multi-anat...

This episode explores a groundbreaking study from Radiology Advances evaluating the use of artificial intelligence as a second reader in screening mammography. Host commentary highlights how the AI-assisted workflow improved cancer detection, reduced radiologist workload, and enhanced reading efficiency, while also emphasizing the importance of thoughtful integration into clinical practice.

In this ai generated episode of the Radiology Advances Podcast, we explore an innovative approach to detecting pulmonary embolism using dual-energy CT without intravenous contrast. Learn how electron density and Z-effective maps could offer a new option for patients with contraindications to contrast media. Pulmonary embolism detection without intravenous contrast using electron density and Z-effe...