🔵 RESEARCH
Trial tests CEUS-MRI hybrid AI to guide early HCC treatment
A clinical trial is testing a hybrid deep learning model that fuses contrast-enhanced ultrasound and MRI to choose between resection and microwave ablation in early hepatocellular carcinoma.
Multimodal AI favours vendors with access to both ultrasound and MRI data streams — GE, Philips, Siemens — over single-modality startups. Clinical superiority over current imaging-only workflows remains unproven until comparative endpoints publish.
🔵 RESEARCH
ML model predicts pregnancy complications in immune-abnormal women
Researchers built and validated a machine learning model that uses first-trimester sonographic features to predict adverse pregnancy outcomes in women with autoimmune abnormalities.
Specialised obstetric AI targeting high-acuity populations opens premium reimbursement pathways for maternal-fetal medicine. Generalisability is the open question — the model likely trained on selected populations at specialised centres and needs external validation across diverse sonographer skill levels.
🔵 RESEARCH
Study refines imaging for immunotherapy response in colorectal cancer
European Radiology study evaluates new qualitative morphological imaging criteria for assessing neoadjuvant immunotherapy response in locally advanced colorectal cancer with dMMR/MSI-H or POLE/POLD1 mutations.
Atypical immunotherapy response patterns make pCR assessment unreliable on conventional criteria, creating space for biomarker-specific AI response tools and companion diagnostic partnerships with immuno-oncology developers.
🔵 RESEARCH
ML standardises first-trimester fetal heart rate for preterm risk
Journal of Ultrasound in Medicine reports a machine learning method that standardises fetal heart rate measurement on first-trimester ultrasound to predict preterm birth.
Early preterm risk stratification is a differentiated feature ultrasound OEMs can productise for maternal-fetal medicine. Adoption depends on whether departments can fit standardised FHR capture into existing OB protocols and surface predictions at the right decision point in the EHR.
Journal of Ultrasound in Medicine · 2026-05-13
Read more → (paywall)
🔵 RESEARCH
Pediatric EM physicians validated on cardiac standstill POCUS
Simulation study assessed pediatric emergency medicine physicians' ability to acquire diagnostic cardiac POCUS images for standstill assessment within 10 seconds, before and after brief training.
Simulation competency is the easy part. Departments still need defined scope-of-practice policies, malpractice clarity, and 24/7 equipment access before physician-performed cardiac POCUS becomes routine in pediatric arrest protocols.
🔵 RESEARCH
SickKids trials AI tool to guide hydronephrosis surgery decisions
Clinical trial integrates a SickKids-developed AI model that uses ultrasound findings to help clinicians decide whether congenital hydronephrosis cases require surgery.
AI is moving into narrow specialist decision workflows previously protected by expert interpretation, pressuring urological imaging incumbents. Sustained adoption hinges on liability frameworks for when AI recommendations influence surgical timing and integration with existing urology scheduling.
🔵 RESEARCH
DL model predicts liver stiffness from standard MRI, no elastography
Multi-site, multi-vendor study developed and validated a deep learning model that regresses liver shear stiffness from standard multiparametric abdominal MRI, bypassing dedicated MR elastography hardware.
A software-only path to liver stiffness threatens the $400M+ elastography hardware market and gives MRI vendors leverage in chronic liver disease workflows. Real-world adoption requires protocol standardisation across MRI fleets and clinician trust versus established FibroScan metrics.
🩺 EDITOR’S TAKE — FOR CLINICIANS
Today's research stack is about replacing specialist judgment with algorithmic decision support — from HCC treatment choice to hydronephrosis surgical timing. Read the methods sections carefully: most lack head-to-head comparisons against your current standard of care.
📊 EDITOR’S TAKE — FOR INVESTORS
Software-only displacement of imaging hardware is the trade to watch. A DL model regressing liver stiffness from standard MRI threatens dedicated elastography vendors, while multimodal AI favours OEMs with both ultrasound and MRI data over single-modality startups.
🏭 EDITOR’S TAKE — FOR INDUSTRY
Multimodal data access is becoming the moat. OEMs spanning ultrasound and MRI gain defensible ground as integrated AI workflows mature, but operational integration — protocol standardisation, liability frameworks, EHR surfacing — remains the bottleneck between research validation and routine deployment.