Question about metric

[Lycus99]

Thanks for your work.
I have a question about the metric used in TransUNet. In the /TransUNet/trainer.py file, I found the metrics defined as follows:

def calculate_metric_percase(pred, gt):
pred[pred > 0] = 1
gt[gt > 0] = 1
if pred.sum() > 0 and gt.sum() > 0:
dice = medpy.metric.binary.dc(pred, gt)
hd95 = medpy.metric.binary.hd95(pred, gt)
return dice, hd95
elif pred.sum() > 0 and gt.sum() == 0:
return 1., 0.
else:
return 0., 0.

I think when pred.sum()==0 and gt.sum()==0, the dice score should be 1.0, not 0.0, and when pred.sum()>0 and gt.sum()==0, the dice score should be calculated by taking the complement of the original pred and gt.

[Kyfafyd]

Hi @Lycus99 , thanks for your interest!
For this metric, we directly adopt it from https://github.com/Beckschen/TransUNet, you can refer to this discussion: Beckschen/TransUNet#39