AI in Mammography Cuts Interval Cancers by 12%

RadAI Slice Newsletter

Weekly Updates in Radiology AI

Good morning, there. AI-supported mammography reduced interval breast cancer diagnoses by 12% in a RCT.

I am struck by the scale and rigor of this Swedish randomized trial, enrolling over 100,000 women and directly measuring outcomes that matter most to both patients and radiology teams. The consistent reduction in advanced and aggressive cancers, plus a 44% drop in radiologist workload, places these findings at the forefront of clinical AI translation. This study stands out for its potential impact on workforce planning and future guideline development.

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

• AI-Supported Mammography Cuts Advanced Breast Cancer in RCT

• CMS Finalizes 2026 Reimbursement Codes for Cardiac CT and AI

• AI Improves Sensitivity and Speed in Lung Cancer CT Screening

• Deep Learning Slashes MRI T2 Imaging Time for Acute Abdomen

• Plus: 3 newly released datasets, 6 FDA approved devices & 4 new papers.

🦾 AI-Supported Mammography Cuts Advanced Breast Cancer in RCT

RadAI Slice: The MASAI trial shows real-world benefit for AI-assisted breast screening.

The details:

• 105,915 women randomized to AI vs standard double reading

• Interval cancer diagnoses dropped 12% with AI (1.55 vs 1.76/1,000)

• Fewer invasive (16%) and aggressive (27%) cancers in the AI group

• 81% of cancers detected at screening in the AI group vs 74% control

• False positive rates unchanged; workload cut by 44%

Key takeaway: The largest RCT to date positions AI not as a radiologist replacement but as a robust aid—delivering earlier cancer detection, reducing advanced case rates, and providing workflow relief as demand for screening grows.

💸 CMS Finalizes 2026 Reimbursement Codes for Cardiac CT and AI

Image from: Radiology Business

RadAI Slice: CMS confirms code and payment changes for CCTA, FFR-CT, and plaque analysis AI.

The details:

• Introduces Category 1 code for AI-enabled coronary plaque analysis in 2026

• Describes new payment policies and operational best practices

• Aims to guide integration and billing for advanced cardiac imaging

• Affects imaging admins, coders, and clinical teams across the U.S.

Key takeaway: U.S. radiology providers should prepare now for substantial clinical and financial shifts in cardiac CT service lines and AI integration—with new billing, workflow, and prior auth processes on the horizon.

🫁 AI Improves Sensitivity and Speed in Lung Cancer CT Screening

RadAI Slice: Georgetown University study validates AI’s impact in lung cancer CT screening.

The details:

• 16 radiologists read 340 low-dose CTs with and without AI

• Sensitivity rose from 0.59 to 0.73 (24.3%) with minimal specificity change

• LROC AUC increased from 0.65 to 0.76

• Reading time dropped from 133s to 115.9s per scan

• Greatest impact seen for small nodules and screenings

Key takeaway: These findings confirm that AI can enhance early lung cancer detection while reducing radiologist workload, especially for challenging small lesions in screening populations.

⏱️ Deep Learning Slashes MRI T2 Imaging Time for Acute Abdomen

RadAI Slice: Prospective trial shows deep learning T2 MRI improves acute abdominal imaging.

The details:

• 70 subjects (healthy and acute abdomen) in a clinical pilot

• SSFE-DLR MRI cut motion artifacts and improved biliary/appendix clarity

• AUC for acute disorders of 0.977–1.0 (vs 0.585–0.953 for standards)

• Scan time reduced; key diagnoses found in vulnerable patients

Key takeaway: Rapid, high-quality DL-enhanced T2 MRI may expand emergency MRI utility for acute abdominal presentations and vulnerable patients, promoting faster, more confident triage.

Emory WMH (January 2026)

Modality: MRI | Focus: Brain, white matter | Task: Segmentation, detection

• Size: 195 scans, 195 patients

• Annotations: Manual WMH segmentations, expert-reviewed

• Institutions: Emory University, Georgia Institute of Technology

• Availability:

  public (Zenodo link)

• Highlight: First diverse, real-world clinical MRI WMH dataset with expert segmentations across 71 scanners.

PMCanalSeg (2026-01-13)

Modality: CBCT | Focus: maxilla, mandible | Task: segmentation

• Size: 191 scans, 191 patients

• Annotations: Dense voxel-level segmentations for pterygopalatine and mandibular canals

• Institutions: Jilin University, Hospital of Stomatology et al.

• Availability:

  public (link)

• Highlight: First dataset with maxillary pterygopalatine canal segmentation in CBCT

PediURF (2026-01-20)

Modality: X-ray | Focus: Forearm (Ulna and Radius), Pediatric | Task: Classification, Detection

• Size: ~10,000 images from 5,265 pediatric patients, each case includes two views

• Annotations: Expert labels for fracture location (proximal, midshaft, distal), case-level classification by radiologists

• Institutions: Shenzhen Children’s Hospital, Dongguan University of Technology, et al.

• Availability:

  Public (https://doi.org/10.6084/m9.figshare.29998954)

• Highlight: First large-scale, public pediatric forearm fracture dataset with dual-view (AP/lateral) images and expert annotation

🏛️ FDA Clearances

• K251934 - FDA clears qXR-Detect for automated chest X-ray abnormality detection, supporting radiologist triage and diagnosis.

• K254186 - Azurion R3.1 receives clearance as an advanced interventional fluoroscopic X-ray system for real-time, high-quality imaging.

• K253023 - Siemens Healthineers gets 510(k) for BIOGRAPH One, a hybrid PET/MR system integrating emission tomography and MRI.

• K252934 - Diagnocat AI gains FDA clearance to assist clinicians with radiology scan analysis and workflow efficiency.

• K254001 - VERITON CT Digital SPECT/CT Series approved for combined anatomical and metabolic imaging for whole-body diagnostics.

• K252579 - Orthoscan TAU MVP Mini C-Arm receives clearance for real-time intra-op MSK fluoroscopy.

• Explore last week's 7 radiology AI FDA approvals.

📄 Fresh Papers

• doi:10.64898/2026.01.25.26344809 - Multi-dataset validation reveals that technical imaging parameters—particularly view type—drive most chest X-ray AI performance disparities.

• doi:10.1016/j.cmpb.2025.109161 - A multicenter study validates a deep learning radiomics nomogram for survival prediction in SCLC, with C-indices up to 0.89 across cohorts.

• doi:10.64898/2026.01.24.26344771 - Transformer-based AI detects critical congenital heart disease from echocardiograms in >54,000 studies, with strong external validation after domain adaptation.

• doi:10.3174/ajnr.A8992 - Prospective clinical validation finds that deep learning–accelerated 3D brain MRI is non-inferior to industry-leading Wave-CAIPI accelerations for detecting intracranial lesions.

• Browse 148 new radiology AI studies from last week.

📰 Everything else in Radiology AI last week

• Combining deep learning with Monte Carlo simulations speeds up patient-specific QA for online adaptive radiation therapy.

• Academic leaders call for new licensure frameworks to regulate clinical generative AI models, proposing a federal digital licensing board.

• Organizational and cultural obstacles hinder integration of AI-driven decision support tools in radiology, according to multi-stakeholder analysis.

• Stanford AI Lab and Rad Partners will collaborate on frameworks for safe scaling and monitoring of imaging AI in clinical practice.

• AI-based analysis of routine CAC CTs can predict future COPD risk, offering prognostic value beyond emphysema quantification.

• Read 17 fresh radiology AI news articles for last week.

That's it for today!

👋 Quick favor: drag this into your Primary tab so you don’t miss next week. Or just hit Reply with one thought.

See you next week.

P.S. We keep building free tools to accelerate your radiology work. What's the most time-consuming pain point in your day that we should help speed up? Reply and share your take so we keep building around you.