🔵 RESEARCH
Cryoablation shown feasible for umbilical endometriosis
European Radiology reports a 17-patient case series on percutaneous image-guided cryoablation for umbilical endometriosis, assessing safety, feasibility and imaging outcomes.
If validated against surgical excision, cryoablation expands interventional radiology's footprint beyond oncology into gynaecology. But with no control arm and a small cohort, recurrence and comparative effectiveness data are needed before this challenges standard care.
🔵 RESEARCH
MonoUNet runs knee cartilage segmentation on portable ultrasound
Researchers published MonoUNet, a compact deep learning model for automated knee cartilage segmentation designed to run on point-of-care ultrasound hardware.
Edge-deployable AI could let portable ultrasound vendors like Butterfly or Clarius offer real-time MSK analysis without cloud fees. The open question is whether segmentation holds up across device brands and non-expert operators in real clinical settings.
Ultrasound in Medicine & Biology · 2026-05-08
Read more →
🔵 RESEARCH
DETECT-AS trial tests AI risk scores for early aortic stenosis
A three-site US study will evaluate a digital biomarker score for precision risk stratification in older adults with aortic sclerosis or mild aortic stenosis.
Following last week's multicentre trial flagging valve disease on routine chest CT, AI-based AS risk tools are accumulating clinical evidence. Whether DETECT-AS beats established scores like STS will determine if digital biomarkers earn defensible reimbursement pathways.
🔵 RESEARCH
Ophthalmic AI pivots from scale to clinical reasoning
NPJ Digital Medicine published a perspective arguing ophthalmic AI should prioritise smarter clinical reasoning architectures over ever-larger models.
If the field accepts this framing, leaner startups with workflow-focused designs could undercut compute-heavy incumbents on deployment cost. Watch for ophthalmic AI companies repositioning marketing away from model size toward clinical integration.
NPJ Digital Medicine · 2026-05-10
Read more → (paywall)
🔵 RESEARCH
Multi-dimensional MRI model predicts stroke functional outcomes
NPJ Digital Medicine published research using multi-dimensional MRI representation and privileged learning to predict functional outcomes in ischaemic stroke patients.
Stroke prognosis AI only helps if it runs inside the treatment window. Most EDs lack the specialised sequences and compute these academic models require — meaningful adoption depends on integration with stroke-centre EHRs and vendor imaging workflows.
NPJ Digital Medicine · 2026-05-10
Read more → (paywall)
🔵 RESEARCH
Deep learning auto-detects sagittal plane in first-trimester NT scans
Journal of Ultrasound in Medicine reports a structure-guided deep learning model that automatically identifies the standard sagittal plane in 3D ultrasound for nuchal translucency assessment.
Automated plane detection lets GE, Philips and other OEMs justify premium pricing on AI-enabled 3D systems for high-volume OB screening. The catch: sonographers must adopt 3D acquisition workflows, and clinic throughput data will decide whether the swap is worth it.
Journal of Ultrasound in Medicine · 2026-05-08
Read more → (paywall)
🔵 RESEARCH
Self-supervised nomogram flags prenatal HLHS risk
Journal of Ultrasound in Medicine published a multimodal nomogram using self-supervised learning to assess prenatal risk of hypoplastic left heart syndrome.
Self-supervised learning helps where labelled data is scarce — exactly the problem for rare congenital defects. But deployment will concentrate at tertiary fetal cardiac centres; specialist AI vendors, not generalist imaging players, are best positioned to commercialise.
Journal of Ultrasound in Medicine · 2026-05-08
Read more → (paywall)
🔵 RESEARCH
Laser-induced acoustic imaging maps nerves and vessels in robotic surgery
Researchers demonstrated laser-induced acoustic imaging to visualise hidden nerves and blood vessels in real time during robotic surgical procedures.
Real-time anatomical guidance could reduce intraoperative injury, but surgical robots are closed ecosystems. Without an Intuitive Surgical or Medtronic partnership and FDA clearance pathway, this stays in the lab for years regardless of clinical promise.
Medical Xpress — Health Research · 2026-05-11
Read more →
🩺 EDITOR’S TAKE — FOR CLINICIANS
Specialist AI is fragmenting into narrow, high-value niches — fetal NT planes, HLHS risk, stroke prognosis, knee POCUS. Watch which tools actually integrate into your scanner workflow rather than demanding offline processing or new acquisition protocols.
📊 EDITOR’S TAKE — FOR INVESTORS
Today's pipeline favours edge-deployable, reasoning-focused models over compute-heavy incumbents. The DETECT-AS trial and Roche's recent PathAI deal signal capital is rewarding clinical evidence and defensible niches, not parameter counts.
🏭 EDITOR’S TAKE — FOR INDUSTRY
OEMs face mounting pressure to embed AI directly on devices — from portable ultrasound to surgical robots. Closed ecosystems like da Vinci and locked imaging platforms will dictate which startups reach commercialisation through partnership rather than standalone clearance.