search

cilix-ai / on-the-fly-guidance

[MICCAI 2024] On-the-Fly Guidance Training for Medical Image Registration. Pre-print available in link below.
https://arxiv.org/abs/2308.15216
Creative Commons Attribution 4.0 International
29 stars 3 forks source link

About VOI_lbls in dice score computation #8

Closed iaOlivia closed 8 months ago

iaOlivia commented 9 months ago

Hi, I want to know how VOI_lbls to be chosen according to different datasest?For example, def dice_IXI(y_pred, y_true): VOI_lbls = [1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18, 20, 21, 22, 23, 25, 26, 27, 28, 29, 30, 31, 32, 34, 36]

why VOI_lbls = [1, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18, 20, 21, 22, 23, 25, 26, 27, 28, 29, 30, 31, 32, 34, 36]? It has any reasons for these values ? Best

yuelinxin commented 9 months ago

Hi @iaOlivia, our datasets are mainly based on the ones used in the Learn2Reg challenge, we used the same labels provided by them so that we can compare our results more easily. These labels are the structures we are most interested in when doing registration.

Additional information can be found at Learn2Reg's website, there you will also find some other datasets: https://learn2reg.grand-challenge.org/Datasets/

If you are interested, you are welcomed to add a new dataset to our work by creating a pull request.

iaOlivia commented 9 months ago

Hi, I check the datasets link you provide to me https://learn2reg.grand-challenge.org/Datasets/ ,can you please tell me how to get the labels used to compute dice?I download OASIS datasets but not found in files where labels might in.

yuelinxin commented 9 months ago

Hi @iaOlivia, I'm not sure the one on the website is fully preprocessed, you can use this one here from TransMorph: https://drive.google.com/uc?export=download&id=1BdEaylMDpeXtyuX5QH8l_Ut4OgenKss4

This is the same with what we used for training our model, just make sure the path names are correct.