Stanford researchers introduce Merlin, a 3D vision-language AI model for interpreting abdominal CT scans, demonstrating strong performance across multiple radiology tasks.
Key Details
- 1Merlin is a 3D vision-language model designed for abdominal CT interpretation.
- 2Evaluated on over 44,000 CT scans from multiple sites covering various anatomies.
- 3Trained on over 6 million CT images, ~2 million diagnosis codes, and 6 million radiology report tokens.
- 4Tested on 752 individual tasks: zero-shot classification, disease risk prediction, cross-modal retrieval, report generation, and 3D organ segmentation.
- 5Achieved 0.741 F1 score for zero-shot classification and AUROC of 0.757 for chronic disease risk prediction over five years.
- 6Model and code are publicly available on GitHub, HuggingFace, and PyPI.
Why It Matters
This foundation model provides a significant step forward in automating CT scan interpretation, potentially alleviating radiologist shortages and expediting workflow. Its public release allows the broader research community to build and benchmark further advances.
Source
AuntMinnie
Related News
•AuntMinnie
Radiology Receives Declining Share of Industry Research Funding
Radiologists received only 1.1% of industry-funded research payments in 2024, with a continuing downward trend.
•AuntMinnie
GPT-4o AI Matches Radiologists in Follow-Up Imaging Recommendations
GPT-4o matched the performance of experienced radiologists and surpassed residents in recommending follow-up imaging from routine radiology reports.
•Cardiovascular Business
AI Leverages Head CTs for Automated Heart Risk Assessments
AI models can turn routine head CT scans into automated cardiovascular risk assessments, expanding the utility of radiology studies.