AI in Radiology
Radiology has the most AI tools of any specialty. Triage of chest X-ray, head CT bleeds and stroke, and screening mammography reading are in routine use in many centres. Report drafting and general-purpose imaging foundation models are still early.
What AI does in Radiology
Flags films likely to show pneumothorax, consolidation, nodules or TB signs so these are read first. Also used in TB screening programmes.
Examples: Qure.ai qXR, Lunit INSIGHT CXR
Alerts the team when a head CT or CT angiogram shows a bleed or large vessel occlusion.
Examples: Aidoc, Qure.ai qER
AI acts as a second reader or sorts normal from suspicious screens. A Swedish randomised trial (MASAI) tested AI-supported reading against standard double reading.
Examples: Lunit INSIGHT MMG, Transpara
Automatic outlines of organs on CT and MRI for volumes, planning and follow-up measurements.
Examples: TotalSegmentator, nnU-Net
Finds nodules on chest CT and compares size with the prior scan.
Examples: Products listed in the ACR AI Central database
Vision-language models write a first draft of a report from the image. Accuracy and safety are under study.
Examples: Research models, not a routine tool
Large models pre-trained on many scans and later adapted to a task with little extra data.
Examples: Research models
routine = in everyday use in many centres · early adoption = in use at some sites, evidence still building · research = studies only
Key ideas for this specialty
New to the basics? Start with the theory lessons.
Tools
Searchable database of imaging AI products with listed regulatory status.
Regulatory status: Database lists status per product
Open platform for imaging AI challenges, datasets and algorithms.
Regulatory status: Not listed
Open-source tool that segments many organs on CT and MRI.
Regulatory status: Not listed
Open-source self-configuring segmentation framework used as a research baseline.
Regulatory status: Not listed
Large free archive of cancer imaging datasets for research.
Regulatory status: Not listed
Indian company with chest X-ray and head CT AI products.
Regulatory status: Not listed
Independent register of radiology AI products with evidence summaries.
Regulatory status: Not listed
A listing here is not an endorsement. Ask any vendor for its CDSCO licence and class before clinical use.
Where to learn
Practical pages on how to choose and monitor AI tools.
See how algorithms are benchmarked on real tasks.
On-demand case-based modules for residents; foundational, advanced, emergency and chest tracks.
Joint five-society checklist for the whole life of an AI tool.
Clear overview by Rajpurkar and Lungren.
Includes diagnosis from chest X-ray and MRI segmentation.
Do it yourself
Install TotalSegmentator, run it on a de-identified or public CT, and view the organ masks. Check three organ volumes against your own judgement of the images.
Open the dataset or tool →Pick a collection on The Cancer Imaging Archive, download a few cases and open them in a free viewer. Note what labels exist and what is missing.
Open the dataset or tool →Open a Grand Challenge task in your sub-specialty. Read how the test set was built and ask whether it matches your patients.
Open the dataset or tool →Use public or properly de-identified data only, and get ethics approval before any project on your own patients.
Papers worth reading
- Developing, Purchasing, Implementing and Monitoring AI Tools in Radiology: Practical Considerations. A Multi-Society Statement Radiology: Artificial Intelligence, 2024
- The Current and Future State of AI Interpretation of Medical Images New England Journal of Medicine, 2023
- Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI) The Lancet Oncology, 2023
- Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI) The Lancet Digital Health, 2025
Societies and guidance
Indian national radiology society; check its site for AI sessions and courses.
AI usage guidance and the AI Central product database.
AI resources and guidance for UK radiology.
AI certificate program and imaging AI education.
Recent updates
Chest X-ray AI did not improve resident accuracy in a prospective crossover study
In Academic Radiology, five radiology residents read 1,861 chest radiographs with four commercial AI tools. Accuracy did not improve overall and fell for pleural effusion and nodules because of false positives, though three of five residents read 6 to 17 seconds faster.
Source: Omnicuris MedshotsFICCI-Praxis report names AI diagnostics as the first large-scale medtech use in India
The report 'AI in MedTech' was released by Union Health Minister J.P. Nadda at India Medical Device 2026. It lists data readiness, regulation and reimbursement as the three gaps between pilots and wide use.
Source: Medical DialoguesHOPPR releases a chest CT foundation model that writes narrative findings
HOPPR's fifth foundation model reads 3D chest CT volumes and describes findings across lung, upper abdomen, heart, mediastinum and soft tissue, including pneumothorax and pulmonary embolism. It is offered through the company's deployment services to radiology departments.
Source: Diagnostic ImagingHarrison.ai releases Harrison.Rad 1.5 for research use
The radiology model drafts narrative reports using the clinical question, patient history and prior studies, and was trained on about 6 million imaging studies. It is not cleared for clinical use and is available for research access only.
Source: AuntMinnieLancet: full MASAI results show AI-supported mammography screening lowered interval cancers by 12%
In the Swedish randomised trial of over 100,000 women, AI-supported screening cut interval cancer diagnoses by 12% with similar false positives (1.5% vs 1.4%). It also found 16% fewer invasive and 27% fewer aggressive-subtype cancers between screens.
Source: ecancerGoogle releases MedGemma 1.5 and MedASR as open models
MedGemma 1.5 4B reads CT, MRI and histopathology images and runs locally, and MedASR is a medical dictation speech model that Google says makes 58% fewer errors on chest X-ray dictation than general models.
Source: Google ResearchLinks checked 3 October 2026. “Not listed” means we could not confirm the detail, not that it is absent. Educational material, not clinical advice.