Weekly AI digest

Three things that mattered

• Deep learning turns routine mammograms into breast cancer risk scores

  A retrospective multicenter cohort in JAMA Network Open (Massachusetts General Hospital / Harvard, 123,091 screening mammograms) reached an AUROC of 0.71, against 0.53 for breast density alone. Women classified as high risk had a 3.8 times higher cancer incidence, opening a path to AI-based personalization of screening intervals.

• Aidoc raises 150 million USD and Abbott clears Ultreon 3.0 on both sides of the Atlantic

  Aidoc closed a 150 million USD Series E co-led by Goldman Sachs Asset Management and NVIDIA Ventures, passing a valuation above 1 billion USD with more than 1,000 health systems as users. In parallel, Abbott obtained simultaneous FDA clearance and CE marking for Ultreon 3.0, its AI-enabled intracoronary OCT system.

• ACR 2026 sets the first US governance standard for imaging AI

  At the ACR Annual Meeting (Washington DC, 3 to 7 May), the ACR and SIIM jointly approved the first Practice Parameter for AI in medical imaging, and launched Assess-AI, presented as the first global quality assurance registry for radiology AI algorithms, with initial results published in JACR.

Peer-reviewed

• Deep learning model for breast cancer risk assessment from screening mammography

  JAMA Network Open · Massachusetts General Hospital / Harvard Medical School, 1 May 2026

• Deep learning radiomics for prediction of symptomatic carotid plaque, with external validation

  American Journal of Neuroradiology · Xuanwu Hospital, Capital Medical University, Beijing, 4 May 2026

• biLSTM recurrent neural network for coronary plaque progression on serial IVUS

  EuroIntervention · MedStar Washington Hospital Center / Inselspital, University of Bern, 4 May 2026

• SpineVLM: a vision-language model for automated spine radiograph reporting

  IEEE Journal of Biomedical and Health Informatics · 4 May 2026

• Machine learning classification of tuberculous, pyogenic and metastatic spondylodiscitis on spine MRI

  BMC Musculoskeletal Disorders · Chiang Mai University, 4 May 2026

• 3D lumbar MRI with deep learning reconstruction: reduced acquisition time at comparable image quality

  American Journal of Neuroradiology · Hospital for Special Surgery / GE HealthCare, 4 May 2026

Industry & regulation

• Abbott obtains FDA clearance and CE marking for the AI-enabled Ultreon 3.0 intracoronary system

  28 April 2026. Next-generation intracoronary OCT imaging with AI for automated plaque analysis, stent implantation guidance and vascular morphology quantification. Simultaneous approval accelerates commercial rollout in North America and Europe.

• Azra AI acquires Thynk Health to strengthen incidental findings detection

  28 April 2026. The acquisition unifies early detection of incidentally found nodules and masses with Azra AI's oncology patient tracking infrastructure, targeting follow-up gaps between radiological discovery and clinical referral.

• Aidoc raises 150 million USD in Series E led by Goldman Sachs and NVIDIA Ventures

  29 April 2026. Co-led by Goldman Sachs Asset Management and NVIDIA Ventures, lifting valuation above 1 billion USD. Funds will advance its clinical AI foundation model, expand health system deployment, and strengthen RIS/PACS integration. More than 1,000 health systems are users.

• Viz.ai partners with the National Rural Health Association to extend AI to rural US settings

  30 April 2026. The initiative includes training programs, technology readiness assessment tools, and facilitated access to Viz.ai solutions for critical access hospitals, addressing higher cardiovascular and stroke mortality tied to diagnostic delays.

• SimonMed Imaging expands its AI portfolio with three new clinical applications

  4 May 2026. Calcium Score+, CT Bone Density (opportunistic AI on routine chest CT), and MR Lumbar Spine+, offered as paid add-ons beyond the base exam fee, illustrating direct AI monetization in ambulatory radiology.

Conferences & voices

The ACR 2026 Annual Meeting (Washington DC, 3 to 7 May) marked two institutional advances for imaging AI: the joint ACR-SIIM Practice Parameter for AI in medical imaging, defining standards for validation, clinical integration and oversight, and the launch of Assess-AI, presented as the first global quality assurance registry for radiology AI, with initial results in the Journal of the American College of Radiology. Woojin Kim, Chief Medical Officer of the ACR Data Science Institute, framed the parameter as a basis for expanding facilitated offerings and defining best practices with all stakeholders. Aidoc CEO Elad Walach argued that by 2030 every complex diagnostic decision should be supported by AI that reasons through findings rather than only flagging them.

Source articles are indexed in PubMed with verified DOIs. Manuscript-stage work is excluded. The full weekly report is produced in French.

Get the weekly digest

One email a week on what matters in radiology and imaging AI. Researched by a practicing radiologist. Free, no vendor sponsorship.

By subscribing you agree to receive the weekly digest by email. Unsubscribe anytime. No sharing with third parties.