Heartflow AI delivers cardiac care breakthrough in major US study

RadAI Slice Newsletter

Weekly Updates in Radiology AI

Good morning, there. Heartflow's AI Plaque Analysis drove major cardiac management changes in 50% of patients.

This landmark, prospective 20,000-patient study shows that integrating AI plaque analysis into CCTA substantially improved care and cardiac event risk. The results offer the strongest evidence yet that imaging AI can personalize risk assessment and directly impact cardiovascular outcomes. As payer and policy shifts now follow, this could redefine the standard for cardiac imaging and management.

Welcome to the very first issue of Radiology AI Slice! We’d love your feedback, so just hit Reply and let us know what you think!

Here's what you need to know about Radiology AI last week:

• AI plaque analysis changes cardiac care in 20,000-patient registry

• Eye-gaze AI system corrects chest x-ray errors in readers

• Medicare moves toward $1,000 payment for coronary CT AI

• Explainable AI outperforms in MRI breast cancer detection

• Plus: 6 FDA approved devices & 4 new papers.

🫀 AI plaque analysis changes cardiac care in 20,000-patient registry

RadAI Slice: Major prospective study shows imaging AI directly improves cardiac care and patient outcomes.

The details:

• DECIDE Registry enrolled ~20,000 US coronary artery disease patients at 30+ sites

• AI-guided analysis prompted treatment changes in more than half of cases

• Mean LDL cholesterol fell 18.7 mg/dL; cardiac event risk dropped by 15% (estimated)

• Significant management shifts occurred even in patients without calcified plaque

Key takeaway: Large-scale real-world evidence shows coronary plaque AI can change care and lower cardiac event risk, propelling AI toward routine management of CAD.

👁️ Eye-gaze AI system corrects chest x-ray errors in readers

RadAI Slice: A novel AI system uses radiologist eye-gaze and reporting data to flag and fix chest x-ray perceptual misses.

The details:

• CoRaX acts as a 'virtual second reader' for chest x-ray interpretation

• Corrected 21–35% of simulated perceptual errors in rigorous evaluations

• Provided diagnostic aid in over 78% of radiologist-AI collaborative cases

• System excelled in missed cardiomegaly findings and potential real-world scenarios

Key takeaway: Collaborative AI leveraging gaze and report data could substantially reduce diagnostic errors, supporting safer, more consistent radiology workflows.

💰 Medicare moves toward $1,000 payment for coronary CT AI

RadAI Slice: CMS is set to bring $1,000 reimbursement to AI coronary plaque analysis, if finalized for 2026.

The details:

• Covers solutions by Cleerly, Elucid, and HeartFlow on CCTA images

• First-ever Category 1 CPT payment code for cardiac CT AI software

• Reflects surging heart disease rates and demand for early AI detection

• Could drive broad clinical adoption and expand commercial prospects

Key takeaway: High-dollar Medicare payment signals major mainstreaming for AI cardiac imaging, setting precedent for wider clinical and payer uptake.

🧠 Explainable AI outperforms in MRI breast cancer detection

RadAI Slice: A new explainable AI model detects and localizes MRI-detected breast cancers, surpassing conventional benchmarks.

The details:

• Model trained and validated on 9,700+ breast MRI exams, plus multicenter external data

• Outperformed binary classifiers in balanced and real-world imbalanced cohorts

• Produced pixel-level heatmaps for transparent tumor localization

• Improved specificity at high sensitivity, supporting streamlined screening

Key takeaway: Robust, transparent AI can improve accuracy and trust in MRI breast cancer detection—key for prospective clinical rollout and radiologist adoption.

🏛️ FDA Clearances

• K251808 - Philips receives 510(k) for its Ingenia and related MR systems for full-body imaging and advanced reconstruction.

• K243227 - ACCUTOME's B-Scan, a diagnostic ultrasound device for detailed imaging, earns FDA clearance.

• K251830 - FUJIFILM Sonosite's LX advanced ultrasound system now cleared, promising improved Doppler and imaging workflows.

• K251169 - GE's Vivid Pioneer ultrasound secures clearance for cardiac assessment using enhanced Doppler technology.

• K251520 - Siemens gains FDA 510(k) for Cios Alpha and Cios Flow, high-end interventional fluoroscopic X-ray systems.

• K251766 - SimBioSys wins clearance for TumorSight Viz, an automated image analysis tool for tumor characterization.

📄 Fresh Papers

• doi:10.1101/2025.07.14.25331522 - A foundation model combining SLO imaging and OCT outperformed prior MSC classifiers using bi-modal retinal features.

• doi:10.12122/j.issn.1673-4254.2025.07.24 - Multi-feature fusion AI classified fetal orientation in intrapartum ultrasound videos with 98.4% accuracy.

• doi:10.1038/s41746-025-01867-w - A review discusses challenges and opportunities for commercializing medical imaging AI, with focus on global impact.

• doi:10.1186/s12903-025-06444-x - ChatGPT-4o correctly diagnosed 78% of complex oral/maxillofacial radiology cases, showing high clinical promise.

📰 Everything else in Radiology AI last week

• EviCore updated its insurance policy to reimburse AI-driven CCTA analysis for cardiac care.

• ACR launches a new committee to address reimbursement challenges for imaging AI in US healthcare.

• A study finds GPT-4o can be inconsistent in imaging tasks, with high accuracy in some cases, but unreliable overall.

• A large survey reports patients perceive doctors using AI as less competent, despite high adoption rates.

• The FDA cleared 211 medical AI devices in the latest update, with over 80% designed for radiology.

That's it for today!

👋 We'd love your thoughts. Just hit Reply and tell us what you want to see in the next Slice!

See you next week.