Stanford researchers unveil Merlin, a foundation AI model that outshines specialist models in analyzing 3D CT scans for diagnostics and disease prediction.
Key Details
- 1Merlin is a vision-language foundation model for 3D abdominal CT analysis, trained on over 15,000 scans, radiology reports, and nearly 1 million diagnosis codes.
- 2The model was evaluated on over 50,000 unseen abdominal CT scans from four hospitals.
- 3On diagnostic coding, Merlin achieved over 81% accuracy across 692 codes and 90% for a subset of 102 codes, outperforming specialist models.
- 4Merlin accurately predicted 5-year risk of chronic diseases from scans 75% of the time, compared to 68% for an existing model.
- 5The model processed both standard and out-of-domain CT scans (chest), matching or surpassing existing tools even outside its training scope.
- 6Researchers hope the model will accelerate clinical workflows and support new biomarker discovery in radiology.
Why It Matters
Merlin's broad capabilities and superior performance signal a shift towards foundation models in medical imaging AI, offering improved diagnostic accuracy, workflow efficiency, and potential for new clinical insights in radiology practice.
Source
EurekAlert
Related News
•EurekAlert
AI Accelerates Radiopharmaceuticals, Boosts Personalized Dosimetry in Cancer
Machine learning is driving advancements in radiopharmaceutical drug discovery and optimizing patient-specific dosimetry for precision cancer therapy.
•EurekAlert
Physicians Overly Trust Erroneous AI, Ignore Contradictory Evidence
Physicians tend to trust incorrect AI advice, even when evidence contradicts it, suggesting risks in clinical decision-making with AI tools.
•EurekAlert
Concerns Raised Over Unverified Datasets in AI Health Prediction Models
A new study finds widely used AI health prediction models are built on datasets with unverifiable origins, raising safety and validity concerns.