RSNA 2025: Radiology AI Goes Clinical

RadAISlice Newsletter

Weekly Updates in Radiology AI

Good morning, there. RSNA 2025 showcased AI tools now embedded in imaging workflows, FDA-cleared for diagnosis, and validated against real outcomes.

The shift at RSNA this year was subtle but deep: AI moved from ‘interesting demos’ to core infrastructure. Vendors aren’t pitching AI—they’re building around it. Radiologists are choosing FDA-cleared tools when they add clarity or speed. Researchers are validating image-derived biomarkers tied to long-term outcomes. And regulators are making AI updates safer to manage. The field is growing up fast—and it’s moving closer to daily clinical relevance.

Which radiology workflow step would benefit most from embedded AI support?

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

• RSNA 2025: AI earns trust in radiology’s core workflow

• Breast ultrasound AI tool may slash unnecessary biopsies

• Multicenter study: AI-based lesion detection shapes MS treatment plans

• Plus: 3 FDA approved devices & 3 new papers.

📊 RSNA 2025: AI earns trust in radiology’s core workflow

RadAISlice: From AI-native scanners to image-derived stress markers, RSNA 2025 made AI feel more like infrastructure than add-on.

The details:

• AI-native platforms: GE, Siemens, and Philips launched end-to-end systems embedding AI from patient positioning to acquisition to reporting. Siemens’ digital twin simulates entire hospital workflows.

• FDA-cleared tools: GE’s Pristina Recon DL won FDA approval and 99% reader preference for DBT clarity. Siemens’ Coronary Cockpit maps coronary disease for PCI planning.

• Explainable AI: New models structured AI outputs to match clinical reasoning steps. LLMs trained on radiology tasks outperformed general ones in summarization.

• AI biomarkers: Johns Hopkins validated adrenal gland volume from chest CTs as a stress marker tied to long-term mortality. Clairity AI risk-stratified breast cancer better than density across 245,000 screenings.

Key takeaway: RSNA 2025 confirmed AI’s real-world impact—from workflow-native design to patient-specific risk insights—and showed radiology is no longer testing AI at the edges but implementing it at the center.

🩺 Breast ultrasound AI tool may slash unnecessary biopsies

RadAISlice: AI-assisted breast ultrasound significantly reduced benign biopsy calls in a prospective multi-reader study.

The details:

• Tested by 14 radiologists on 271 lesions in 253 women

• 60.6% specificity, sensitivity 90.5%

• Misclassified 6 malignancies—mostly with other imaging flags

• Reduces unnecessary biopsies, lowers patient anxiety and costs

Key takeaway: AI can help radiologists cut benign biopsies in breast ultrasound, easing operational burden and improving patient care without sacrificing safety.

🔍 Multicenter study: AI-based lesion detection shapes MS treatment plans

RadAISlice: Integration of AI in MS follow-up MRIs led to more detected new lesions and altered management decisions.

The details:

• Retrospective cohort: 83 MS patients, mean age 49

• AI-aided MRI readings found new lesions in 30% vs 15% for conventional

• 33.7% detected enlarging lesions with AI, none with standard read

• 10.8% of treatment plans changed after using AI findings

Key takeaway: AI-complemented MRI reading in MS increases detection of disease activity and directly impacts patient management decisions.

🏛️ FDA Clearances

• K251822 - MAGNETOM Free.Max and Free.Star MRI systems cleared, aiming for accessible high-quality MRI at reduced field strength.

• K253495 - syngo.MR Applications (VB80) offers advanced MRI processing tools for radiology image enhancement and analysis.

• K253487 - ACUSON Maple Select Diagnostic Ultrasound System cleared for real-time, multi-application ultrasound imaging in clinical settings.

• Explore last week's 11 radiology AI FDA approvals.

📄 Fresh Papers

• doi:10.1007/s00330-025-12185-w - Croatia's national lung screening program combines LDCT, AI-assisted volumetrics, and GP-centered recruiting; over 70K scans in 5 years.

• doi:10.1002/uog.70146 - First-trimester placental ultrasound with AI segmentation improves fetal growth restriction prediction (AUC 0.79, 3500 pregnancies).

• doi:10.1038/s41598-025-28221-0 - Dual-pathway AI combines brain MRI and clinical data for Alzheimer's prediction, validated on 42k+ images, with competitive accuracy.

• Browse 170 new radiology AI studies from last week.

📰 Everything else in Radiology AI last week

• Survey finds 75%+ of radiology organizations using AI lack clear, quantified ROI data, though most report positive or neutral impressions.

• UCLA-led studies at SABCS 2025 show AI pathology and blood biomarkers improve breast cancer risk stratification and treatment decisions.

• Dr. Paul Chang discusses maturity of radiology AI, the potential of foundation models, and emergence of agentic AI at RSNA 2025.

• Read 13 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.