RadAI Slice Newsletter Weekly Updates in Radiology AI |
Good morning, there. Integrating MRI with other imaging modalities in AI boosts breast cancer risk prediction AUROC from 0.899 to 0.94. I’m struck by how much the addition of MRI to AI models improved 5-year cancer risk prediction for women in NYU's massive multimodal study. This finding feels especially relevant for radiology, as it supports targeted screening and genuinely moves the field toward precision medicine. It also highlights the value of combining modalities and clinical data in practical, large-scale workflows. How would adding multi-modality AI risk stratification affect your screening protocols?
Here's what you need to know about Radiology AI last week: MRI Elevates Multimodal AI for Breast Cancer Risk and Detection Study: <30% of Radiology AI Devices Have Clinical Testing Pre-FDA Clairity Raises $43M for AI Mammogram-Based Risk Prediction AI Repurposes CTs for Osteoporosis Screening, $2.5B Savings Possible Plus: 2 newly released datasets & 4 new papers.
|
Which AI tool do you use most often in your daily work? |
🧬 MRI Elevates Multimodal AI for Breast Cancer Risk and Detection RadAI Slice: NYU researchers show MRI enhances AI-based breast cancer risk prediction and detection. The details: Model trained on 1.3M exams from 274,388 women across multiple modalities Separate test cohort of 1,944 women, 18,201 exams Adding MRI increased AUROC for 5-year risk from 0.899 to 0.94 Supports more accurate high-risk identification and targeted screening
Key takeaway: Integrating MRI with mammography, DBT, and ultrasound in AI models can elevate precision risk stratification and early detection, shaping future breast cancer screening guidelines. |
📜 Study: <30% of Radiology AI Devices Have Clinical Testing Pre-FDA  Image from: Radiology Business RadAI Slice: A review in JAMA reveals most FDA-cleared radiology AI tools lack clinical testing evidence. The details: Review of ~1,000 FDA-cleared AI/ML devices since 1995 723 radiology-focused devices analyzed <30% had any clinical testing before clearance Prospective trials were rare for FDA approval
Key takeaway: Radiology AI adoption remains ahead of validation, highlighting urgent regulatory gaps and the need for standardized clinical evidence to ensure patient safety. |
💸 Clairity Raises $43M for AI Mammogram-Based Risk Prediction  Image from: Radiology Business RadAI Slice: Clairity lands major investment to scale its FDA-cleared mammography AI tool. The details: Series B: $43M to expand access to Clairity Breast Predicts 5-year breast cancer risk from routine mammograms Addresses early detection and risk stratification Funded by Ace Global, Santé Ventures, BCRF
Key takeaway: Strong funding signals confidence in AI risk tools with regulatory approval; adoption could shift prevention and workflow strategies in breast imaging. |
🧮 AI Repurposes CTs for Osteoporosis Screening, $2.5B Savings Possible RadAI Slice: AI models can screen for osteoporosis using routine CT scans, aiding population health. The details: Analyzed 538K CTs from 283,499 patients for bone loss indicators Adjusts for age, gender, ethnicity to map population trends Could double osteoporosis screening rates nationwide Projected $2.5B annual savings in Medicare
Key takeaway: Opportunistic AI screening using existing CT data could uncover millions of undiagnosed osteoporosis cases and drive early intervention, expanding radiologist impact. |
RUS-HUS-LumbarSpine (2025-xx-xx) Modality: US, CT | Focus: Lumbar spine, musculoskeletal | Task: 3D reconstruction, segmentation Size: 223 handheld US scans, 375 robotic US scans, 63 healthy volunteers, paired CT; ~6091 annotated US frames Annotations: Manual bone surface labels, 3D CT segmentations (vertebrae L1-S1); 6091 expert-annotated US images, STL surface models Institutions: Balgrist University Hospital/University of Zurich, KU Leuven Availability: Highlight: Largest paired robot-assisted and handheld lumbar spine ultrasound dataset with CT ground truth, includes synchronized position tracking and segmentations
|
COde (Casangels Oro-dental) (2025-06) Modality: X-ray, Photos | Focus: Teeth, Oral cavity | Task: Classification, Report generation Size: 8775 checkups, 4800 patients, 50000 photos, 8056 radiographs Annotations: Diagnosis labels for 120+ oro-dental diseases, validated by experts; full diagnostic text reports Institutions: Suzhou Doctor Dental Clinic, South China University of Technology et al. Availability: Highlight: Large-scale bilingual multimodal dental dataset with paired images and textual records for VLMs.
|
📄 Fresh Papers doi:10.1101/2025.03.28.25324832 - Automated Cobb angle estimation in 33,889 UK Biobank MRIs shows scoliosis prevalence much higher than previously reported. doi:10.1007/s00247-025-06403-2 - Fully automated fetal biometry for 3D T2 MRI accurately quantifies brain growth with real-time clinical reporting. doi:10.1038/s41746-025-02018-x - GlioSurv transformer integrates MRI/clinical/molecular/treatment data for personalized glioma survival prediction, validating across multiple cohorts. doi:10.1101/2025.11.07.25339752 - Machine learning-assisted phenotypic imputation with bias-corrected GWAS lets MRI-derived organ volumes drive scalable genetic discovery in 450,000 UK Biobank cases. Browse 165 new radiology AI studies from last week.
|
That's it for today! Before you go we’d love to know what you thought of today's newsletter to help us improve the RadAI Slice experience for you. |
|
👋 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. |
|
