AI in Radiology Quiz

Answers and explanations appear as you take the quiz.

1. What operation lets a convolutional neural network detect local image patterns in radiology images?
   • It applies Bayes theorem to patient records
   • Convolutions apply learned filters across neighboring pixels or patches
   • It converts images into a text transcript
   • It uses a spreadsheet formula
2. What is a possible worklist use of a radiology AI model?
   • Replace the radiologist’s final interpretation
   • Guarantee that no urgent case is missed
   • Flag a study for earlier review when a defined finding is suspected
   • Make a diagnosis for every patient without review
3. What does 'domain shift' refer to in radiology AI?
   • Changing the hospital's parking lot
   • Switching from MRI to ultrasound for fun
   • A type of image format
   • A difference between training and deployment data that can change model performance
4. Which architecture is commonly used to outline (segment) a tumor pixel by pixel?
   • U-Net
   • SMILES
   • XGBoost
   • JPEG
5. What was the ScreenTrustCAD study design for screening mammograms?
   • A retrospective study of chest CT only
   • A prospective paired-reader study comparing interpretations from two radiologists and AI CAD
   • A study comparing image file sizes across scanners
   • A trial of autonomous treatment planning
6. Which metrics are commonly used to evaluate a radiology AI model?
   • Horsepower and torque
   • Word count and font size
   • Sensitivity, specificity, and AUC
   • Battery life
7. What does FDA authorization of an AI radiology device establish?
   • That it is accurate for every possible clinical use
   • That it replaces a radiologist’s judgment
   • That it needs no monitoring after deployment
   • That the device met applicable premarket requirements for its intended use
8. How many women were included in the 2025 ScreenTrustCAD human-AI interaction analysis?
   • 54,991 women
   • 55,581 women
   • 5,499 women
   • 549,910 women