AI Lung Nodule Detection Quiz

Answers and explanations appear as you take the quiz.

1. In the National Lung Screening Trial, low-dose CT was compared with which alternative?
   • No screening at all
   • MRI
   • Sputum testing
   • Chest X-ray
2. Which age range did the 2021 USPSTF recommendation set for annual lung cancer screening?
   • 40 to 70
   • 50 to 80
   • 55 to 74
   • 60 to 90
3. A 9 mm nodule turns up by chance on a CT done for abdominal pain. According to the guide, which guidelines usually govern its follow-up?
   • Fleischner Society guidelines
   • Lung-RADS
   • KDIGO criteria
   • NEWS2
4. What makes Sybil, from MIT and Massachusetts General Hospital, different from a standard nodule detector?
   • It performs biopsies robotically
   • It only works on chest X-rays
   • It estimates future lung cancer risk from a single low-dose CT, even with no suspicious nodule visible
   • It replaces the need for radiologists to read scans
5. Why are nodule detectors often evaluated with FROC curves?
   • They show sensitivity against the average false positives per scan, which fits scans that can hold several nodules
   • They measure radiation dose
   • They rank vendors by price
   • They report how long each scan takes
6. What does a nodule's volume doubling time help clinicians judge?
   • The patient's lung capacity
   • How fast the nodule is growing, with faster growth suggesting cancer
   • Which scanner produced the image
   • Whether contrast dye was used
7. Why does the guide say it is a misconception that AI diagnoses lung cancer?
   • AI cannot see ground-glass nodules
   • AI only works on patients who smoke
   • The FDA bans AI in lung imaging
   • AI flags nodules and estimates risk, but only tissue sampling confirms cancer
8. A detector that did well at one hospital performs worse at another. Which explanation does the guide give?
   • The second hospital's patients have larger lungs
   • Nodules look different in different countries
   • Differences in slice thickness, reconstruction kernel, dose and scanner vendor change what the model sees
   • The model forgets its training over time