Weekly AI digest

Three things that mattered

• Accuracy claims and ads raise willingness to pay for AI mammography

  A randomized JACR survey from Brown University (2,534 women aged 40 and over) shows that information about AI accuracy and advertising significantly increase out-of-pocket willingness to pay for AI-assisted mammography interpretation, up to 26.5% when an ad is shown versus 7% when a high error rate is presented. With no CPT code yet for AI-assisted screening mammography, the findings bear directly on billing policy.

• A wave of US state AI legislation reaches radiology

  The ACR is tracking seven bills across five states (California AB 2575, Iowa HSB 766, Alabama SB 63, Minnesota SF 1856, New Hampshire HB 1406) that aim to govern AI in prior authorization and insurance claims, and submitted radiology's perspective to the White House AI Action Plan.

• The "Bixonimania" affair exposes medical AI hallucination risk

  A University of Gothenburg team posted two openly fraudulent preprints describing a fictional eye disease, "Bixonimania." Chatbots including ChatGPT, Gemini, and Copilot then presented the condition as real and advised patients to see specialists, a stark illustration of AI hallucination risk and the dangers of unvetted preprint proliferation.

Peer-reviewed

• AI accuracy and advertising influence women's willingness to pay for AI-assisted mammography

  Journal of the American College of Radiology · Brown University, Providence, 4 April 2026

• Contextual classification of radiology report sentences using synthetic data: a development and validation study

  Journal of Medical Internet Research · 4 to 11 April 2026

Industry & regulation

• GE HealthCare gains FDA 510(k) clearance for True Definition DL CT reconstruction

  2 April 2026. A deep-learning CT image reconstruction solution that improves spatial resolution and suppresses artifacts without added acquisition time or radiation dose, producing a 1024 display matrix; compatible with Revolution Apex and Revolution Vibe platforms.

• Philips launches Verida, an AI-powered spectral detector CT, in Japan

  Launch announced for 17 April 2026. Verida embeds AI across acquisition and reconstruction to deliver spectral and conventional images simultaneously, with up to 45% energy savings, reconstruction at 145 images per second, and capacity for up to 270 exams per day. First shown globally at RSNA 2025.

Conferences & voices

SIIM opened registration for SIIM26, its annual imaging informatics conference, held 10 to 12 June 2026 at the David L. Lawrence Convention Center in Pittsburgh, with the theme "Building Bridges in Imaging Informatics" and more than 150 sessions; early-bird pricing ran through 27 April. On community voices, Eric Topol (Scripps Research Translational Institute) warned about AI-driven medical misinformation after chatbots amplified the Bixonimania affair, and Pranav Rajpurkar (Harvard Medical School, a2z Radiology AI) commented on recent advances in radiology AI on LinkedIn.

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.