leeh43
commented
1 year ago Hi, thank you for your interest to our work! To segment abdominal CT images with multiple class (e.g., spleen, kidney, liver, ...etc), we define the number of classes (out_classes) that we want to segment. For instance, FLARE dataset have four organs labeled, so the number of classes will be 5 (number of organs + background).
For calculating the Dice score, we have leveraged a built-in function in MONAI (dice_metric) to compute the dice score across all categories. However, if you want to output the dice score of each category, you may need to use the dice_metric function for each prediction channel. You can first obtain the one-hot prediction of each channel and evaluation the overlapping region with the ground truth (also in one-hot setting). If you are still not familiar with it, I can send you my own dice score function.
ASmallWhit
CamillerFerros
SanmAIGO
Hello, thank you for your work. I want to ask how to segment abdominal CT images with multiple categories, and how to output the dice score of each category at the same time? I don't think this is reflected in your code. Thank you very much