Dice Coefficient
The most commonly used per-sample evaluation metric in medical image segmentation, measuring the degree of overlap between a predicted mask and ground-truth annotation. It is sensitive to overlap of large regions but tends to give overly optimistic scores for small targets or boundary regions.
Timeline (last 90 days)
The Dice coefficient measures the degree of overlap between the predicted mask and the ground-truth annotation, ranging from 0 to 1, with the formula being twice the intersection divided by the sum of the total number of pixels (or voxels) of the two; it is sensitive to overlap in large-area regions, but tends to give overly optimistic scores on small targets or boundary regions.
The Dice coefficient and the Hausdorff distance are both per-sample metrics, which require the pairing of 'this image corresponds to this annotation' before they can be computed, cannot characterize the distributional properties of the prediction set as a whole, and cannot be computed directly on unlabeled data.
All Facts (2)
The Dice coefficient measures the degree of overlap between the predicted mask and the ground-truth annotation, ranging from 0 to 1, with the formula being twice the intersection divided by the sum of the total number of pixels (or voxels) of the two; it is sensitive to overlap in large-area regions, but tends to give overly optimistic scores on small targets or boundary regions.
50%UnverifiedThe Dice coefficient and the Hausdorff distance are both per-sample metrics, which require the pairing of 'this image corresponds to this annotation' before they can be computed, cannot characterize the distributional properties of the prediction set as a whole, and cannot be computed directly on unlabeled data.
50%