Back to all issues
Issue #8
September 9, 2025

AI brain platform triples UK stroke patient independence

PLUS: Agentic AI tools automate complex medical imaging workflows

RadAI Slice Newsletter

Weekly Updates in Radiology AI

Good morning, there. NHS England’s AI stroke platform tripled functional independence rates.

Efficient AI-driven triage in stroke care dramatically reduced time to treatment, improved functional outcomes, and increased advanced procedure rates nationwide. This broad-scale, real-world impact signals a new era for AI in imaging-driven clinical workflows, highlighting potential benefits for both patients and radiologists.


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

  • NHS AI Stroke Platform Triples Functional Independence

  • Agentic AI Bridges Imaging Interoperability Gaps

  • AI Model Validated for Diverse Lung Cancer Risk Prediction

  • AI Breast Imaging Yields High NPV but Raises Recalls

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

🚀 Product Update

X-ray Interpreter now supports multiple image uploads 📤. Plus, our new Auto Select feature ⚡ can automatically pick key images from a full DICOMDIR folder, saving you time and effort.

👉 Try it now

LATEST DEVELOPMENTS

🧠 NHS AI Stroke Platform Triples Functional Independence

🧠 NHS AI Stroke Platform Triples Functional Independence

Image from: Health Imaging

RadAI Slice: A national AI brain imaging deployment has transformed stroke recovery and care efficiency.

The details:

  • Brainomix 360 used at all NHS stroke centers since summer 2023

  • Over 60,000 patient cases analyzed using CT, CTA, CTP, and MRI

  • Time to treatment fell from 140 to 79 minutes post-implementation

  • Functional independence jumped from 16% to 48% of patients

  • Mechanical thrombectomy procedures rose by 50%

Key takeaway: Robust clinical implementation of imaging AI can drive major real-world improvements in patient outcomes, operational efficiency, and advanced care adoption across an entire health system.

Read the full NHS AI stroke impact story

🤖 Agentic AI Bridges Imaging Interoperability Gaps

RadAI Slice: Agentic AI is emerging as a solution for imaging workflow integration across platforms.

The details:

  • Agentic AI acts as credentialed users to automate procedures

  • Supports requisition intake, exam retrieval, reminders, and follow-up

  • Combines LLMs, vision, and rules for robust, auditable performance

  • Reduces manual touches and repeat scans, optimizes scheduling

Key takeaway: Agentic AI delivers true workflow automation and resilience where standards-based integration still fails, enabling next-level efficiency and safety for radiology teams.

🫁 AI Model Validated for Diverse Lung Cancer Risk Prediction

RadAI Slice: Sybil AI was validated in a diverse, predominantly Black safety-net hospital population.

The details:

  • Validated on over 2,000 CTs at UI Health (Chicago), 2014–2024

  • 62% Non-Hispanic Black, 13% Hispanic, high social diversity

  • AUC stayed strong through 6-year risk window, even in subgroups

  • Consortium advancing to clinical trial for workflow integration

Key takeaway: Equitable screening AI validated in underserved populations signals critical progress against disparities in lung cancer outcomes.

🩺 AI Breast Imaging Yields High NPV but Raises Recalls

RadAI Slice: AI-driven breast screening safely rules out negatives but triggers more recalls than radiologists.

The details:

  • Transpara AI compared to 11 breast radiologists on >30,000 cases

  • NPVs for both: 99.8–99.9%; AI sensitivity up to 94% at higher thresholds

  • AI recall rate up to 41.8% vs <7.2% for radiologists

  • Urgent need for strategies to limit AI-driven false-positives

Key takeaway: AI can streamline breast screening by ruling out negatives, but clinical gains depend on managing increased recall and false-positives.

NEW DATASETS

HECKTOR2025 (2025-09-03)

Modality: PET/CT | Focus: Head and neck | Task: Segmentation, prognosis prediction

  • Size: 1123 scans, 1123 patients

  • Annotations: Manual tumor (primary and lymph node) segmentations, radiotherapy dose maps, clinical outcome and biomarker data

  • Institutions: Mohamed bin Zayed University of AI, MD Anderson Cancer Center et al.

  • Availability:

  • Highlight: Largest publicly available head and neck PET/CT dataset with multimodal annotations and long-term outcome data from 10 centers

African Breast Imaging Dataset (ABID) (2025-08-28)

Modality: Mammogram, US | Focus: Breast | Task: Segmentation, Classification

  • Size: ~400 US scans, ~200 mammograms; ~100 patients, longitudinal

  • Annotations: BI-RADS scores, segmentation masks, breast composition, lesion characteristics, multi-radiologist labels

  • Institutions: Ernest Cook University, MAI Lab et al.

  • Availability:

    Public (planned), medRxiv link

  • Highlight: First open, longitudinal African dataset with both POC and conventional breast imaging, expert labels.

CRL-2025 Atlas (2025-08-28)

Modality: MRI (T2w, DTI) | Focus: Fetal brain | Task: Segmentation, Parcellation

  • Size: 194 MRI scans, 160 fetuses (21-37 weeks gestation)

  • Annotations: Detailed tissue segmentation (36 labels), regional parcellation (126 labels), transient white matter compartments to 31 weeks

  • Institutions: Boston Children’s Hospital, University of California Irvine, et al.

  • Availability:

  • Highlight: High-resolution 4D spatiotemporal fetal brain atlas with unique fine-grained segmentations and open-source tools.

QUICK HITS

🏛️ FDA Clearances

  • K251983 - Brainomix 360 Triage Stroke: FDA-cleared AI triage tool for brain imaging stroke findings, enabling rapid clinical decisions.

  • K251484 - CT:VQ by 4DMedical: FDA-cleared CT-based ventilation imaging tool for detailed lung function assessment.

  • K252526 - Rapid DeltaFuse by iSchemaView: FDA-cleared image processing for efficient radiology decision support.

  • K250947 - VistaSoft 4.0 and VisionX 4.0: FDA-cleared radiology automation software for improved diagnostic image analysis.

  • K250788 - Definium Tempo Select by GE: Stationary X-ray system for diagnostic medical imaging.

  • K251106 - Sonosite LX and PX: FDA-cleared ultrasound imaging systems for real-time, multi-organ clinical assessment.

  • Explore last week's 8 radiology AI FDA approvals.

📄 Fresh Papers

  • doi:10.1016/j.artmed.2025.103254 - LoRA-PT offers efficient fine-tuning of transformer UNETR, improving hippocampus segmentation in scarce data settings.

  • doi:10.1007/s00259-025-07504-8 - AI-assisted quantification of PET/CT metabolic response predicts survival in metastatic uveal melanoma on tebentafusp therapy.

  • doi:10.1038/s41551-025-01497-3 - MedSegX, trained on MedSegDB, achieves state-of-the-art, generalist, open-world medical segmentation—robustly handling OOD data.

  • doi:10.1101/2024.10.17.24315675 - YOLOv8-based deep learning model detects early asymptomatic carotid plaques at population scale, improving risk stratification in CV disease.

  • Browse 205 new radiology AI studies from last week.

📰 Everything else in Radiology AI 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.

⭐⭐⭐⭐⭐ Nailed it
⭐⭐⭐ Average
⭐ Fail

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

See you next week.

Ready to Sharpen Your Edge?

Join hundreds of your peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.