Explainability of Dice Metric on testing dataset

[HripsimeS]

Hello. I evaluated my 3D Unet model on my testing dataset for 5 original images with their masks. I received good score for dice_metric(y_pred, y), but to be total transparent with results I need to check with you some details. I used dice_metric.get_buffer() to check all calculated scores and I received the following results below. I believe 5 rows are the number of my 5 images/masks. But what are the 4 columns exactly, what they are representing?

([[0.9041, 0.9717, 0.9807, 0.9075],
[0.9443, 0.9552, 0.9536, 0.9376],
[0.9721, 0.9837, 0.9757, 0.9751],
[0.9434, 0.9579, 0.9643, 0.9585],
[0.9525, 0.9726, 0.9845, 0.9782]])

[HripsimeS]

I checked on Sleep dataset and results were represented with one column, it means it is a number of classes excluding background.
I am closing the ticket :)