The RADAR vision-language AI model shows expert-level diagnostic ability for abdominal CT across multiple diseases and clinical settings.
Key Details
- 1RADAR is a vision-language AI model trained on 424,911 abdominal CT exams with over 1.5 million image-text and 15 million anatomy-specific pairs.
- 2It breaks CT scans down by anatomical structure and links each to radiology report text using contrastive learning.
- 3RADAR achieved a mean AUC of 0.913 across 146 abdominal CT findings, outperforming the best existing model (AUC 0.776).
- 4In over 27,000 emergency CT cases, RADAR had an AUC of 0.904, despite not being specifically trained for emergencies.
- 5In tests from eight external centers, RADAR maintained high accuracy (AUC 0.895), demonstrating strong generalizability.
Why It Matters
These results suggest that generalist, expert-level AI systems like RADAR can provide reliable, accurate, and scalable diagnostic support for complex CT imaging tasks. This could significantly impact radiology workflows, improve diagnostic consistency, and aid clinicians across diverse clinical environments.
Source
EurekAlert
Related News
•EurekAlert
UVA's $4M NIH Grant Backs AI to Optimize Cardiac MRI Therapies
UVA secures $4 million NIH funding to develop GenStrain, an AI model using MRI data to personalize cardiac pacing therapies.
•EurekAlert
Machine Learning Models Predict Cancer Risk with Routine Health Data
Researchers developed machine learning models using routinely collected health data to predict individual risk for 20 common cancers, outperforming traditional methods for several cancer types.
•EurekAlert
AI Deep Learning Model Enhances Echocardiogram RV Assessment and Prognosis
Researchers developed DROID-RV, a deep learning tool that automates right ventricular assessment from echocardiograms, showing strong prognostic links with cardiovascular outcomes.