Dice Coefficient and Segmentation Metrics Quiz

Answers and explanations appear as you take the quiz.

1. Which counts appear in the binary Dice formula's denominator?
   • TP plus FP plus FN
   • TP plus TN plus FP plus FN
   • 2TP plus FP plus FN
   • TP plus FN
2. A Dice score of one for nonempty masks indicates what?
   • The prediction can miss reference foreground without reducing the score
   • Predicted and reference masks match exactly
   • Equal volumes suffice even if the regions occupy different locations
   • Only the fraction of correctly labeled background pixels matters
3. What does IoU measure?
   • Reference region size divided by predicted region size
   • 2TP divided by 2TP plus FP plus FN
   • Mean distance between predicted and reference boundaries
   • Intersection divided by union
4. Which metric is especially sensitive to one distant boundary error?
   • Hausdorff distance
   • Mean surface distance between boundaries
   • Intersection over union
   • Average symmetric surface distance
5. How can averaging per-case Dice differ from pooling all pixels first?
   • Pooling pixels gives every case equal influence
   • Per-case averaging weights larger masks more heavily
   • Per-case averaging weights cases equally; pooling weights them by their Dice denominators
   • Macro averaging necessarily pools all foreground pixels before scoring
6. Why document how background is handled in multiclass Dice?
   • Background inclusion necessarily leaves a pooled score unchanged
   • Background pixels can dominate counts and alter the summary
   • Excluding background always lowers every foreground-class score
   • Background handling affects boundary distance but never overlap
7. Why use physical spacing for boundary distances in medical scans?
   • Distances measured in voxels have the same physical scale across scans
   • Dice should be multiplied by voxel spacing to become physical
   • Using spacing makes all scans have identical resolution
   • Voxel dimensions may represent different physical distances
8. Which conclusion is warranted by high average Dice alone?
   • Masks overlap well on average under the stated calculation
   • Every individual case has perfect overlap
   • Boundary accuracy is necessarily high
   • The annotations and clinical outcomes are necessarily correct