🔵 RESEARCH
Pediatric liver MRI study compares GRASP-VIBE and CAIPIRINHA-VIBE
European Radiology published a retrospective study of 70 children comparing free-breathing GRASP-VIBE against breath-hold CAIPIRINHA-VIBE for dynamic contrast-enhanced liver MRI at 3T.
Sequence choice in pediatric liver MRI affects diagnostic confidence and patient cooperation. But with a 70-patient retrospective design, clinicians should reserve protocol changes until blinding methods, inter-rater reliability, and diagnostic-accuracy endpoints are reviewed in full.
🔵 RESEARCH
Prospective study tests multiparametric MRI for non-mass breast lesions
A prospective European Radiology study evaluated multiparametric breast MRI combining DCE and DWI to differentiate benign from malignant non-mass enhancement lesions.
If specificity and biopsy-upgrade rates hold up against histopathology, premium MRI vendors gain ammunition for advanced protocol upsell, and breast AI players like iCAD face pressure to move beyond mass lesions into harder NME territory. Single-center generalizability remains the open question.
🔵 RESEARCH
CMU-Cleveland Clinic AI reads cardiac MRI without manual labels
Carnegie Mellon and Cleveland Clinic researchers built a self-supervised AI system for cardiac MRI interpretation that they report outperforms general models by 35%.
If the result holds against supervised baselines on multicenter data, the labeled-dataset moat protecting Ultromics, HeartFlow and Eko weakens materially. Watch whether Cleveland Clinic commercializes directly — a signal that providers are moving to capture AI revenue rather than license it out.
Medical Xpress — Health Research · 2026-05-21
Read more →
🔵 RESEARCH
Meta-analysis backs AI-assisted POCUS for fracture detection
A systematic review and meta-analysis in The Ultrasound Journal assessed AI diagnostic performance for fracture detection using point-of-care ultrasound.
Following this week's FAST-exam meta-analysis, evidence is building for AI-augmented POCUS as a radiation-free X-ray alternative — a tailwind for Butterfly and Sonosite. But ED adoption needs credentialing protocols and EHR-integrated workflows that most departments aren't staffed to deliver yet.
🔵 RESEARCH
JUM publishes commentary on deep learning for skin lesion ultrasound
The Journal of Ultrasound in Medicine ran a comment on a prior CNN-based deep learning paper classifying skin lesions from high-frequency ultrasound imaging.
Sustained peer engagement signals that high-frequency ultrasound is gaining traction as an AI-amenable modality for dermatologic imaging — a niche that could open competitive space adjacent to dermoscopy-focused AI.
Journal of Ultrasound in Medicine · 2026-05-20
Read more → (paywall)
🩺 EDITOR’S TAKE — FOR CLINICIANS
Today's research tilts toward AI that reduces dependence on labeled data and manual protocols — from cardiac MRI to POCUS fracture detection. Treat headline accuracy gains as hypothesis-generating until multicenter validation and blinded reader studies land.
📊 EDITOR’S TAKE — FOR INVESTORS
The CMU-Cleveland Clinic self-supervised cardiac MRI work erodes the labeled-dataset moat that has underpinned valuations at Ultromics, HeartFlow, and Eko. Expect compressed time-to-market for academic spinouts and harder questions about defensibility in cardiac AI rounds.
🏭 EDITOR’S TAKE — FOR INDUSTRY
Evidence is stacking up for protocol-specific MRI workflows (breast NME, pediatric liver) and POCUS-as-X-ray-substitute — both expand the addressable surface for OEM software differentiation. But operational gaps in ED credentialing and EHR-integrated POCUS AI remain the real adoption bottleneck.