search

lihaoliu-cambridge / unsupervised-medical-image-segmentation

Code for "Contrastive Registration for Unsupervised Medical Image Segmentation".
63 stars 14 forks source link

label fusion in mindboggle 101 dataset #1

Closed zhangliutong closed 5 months ago

zhangliutong commented 4 years ago

I wonder if you follow the FAIM to use the label_list_25 to calculate the dice, and then fuse that into 5 regions to get the final result.

lihaoliu-cambridge commented 4 years ago

Yes, I followed this setting.

On Thu, 7 May 2020 at 11:23 AM, zhangliutong notifications@github.com wrote:

I wonder if you follow the FAIM to use the label_list_25 to calculate the dice, and then fuse that into 5 regions to get the final result.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/liulihao-cuhk/Probabilistic-Multilayer-Regularization-Network/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AESZCO4Z3GBCHGTMVI234IDRQKDZPANCNFSM4M3GKNLA .

zhangliutong commented 4 years ago

I am confused that why not first fuse the regions into 5 ROIs, and then calculate the ROI dice.

lihaoliu-cambridge commented 4 years ago

Actually, you can.

Since in the training process you didn’t use any segmentation mask, you can also use the fuse-first segmentation mask for evaluation as long as you compare with other methods on this kind of segmentation mask.

On 7 May 2020, at 12:18 PM, zhangliutong notifications@github.com wrote:

I am confused that why not first fuse the regions into 5 ROIs, and then calculate the ROI dice.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/liulihao-cuhk/Probabilistic-Multilayer-Regularization-Network/issues/1#issuecomment-625192792, or unsubscribe https://github.com/notifications/unsubscribe-auth/AESZCO22OMSUITLHAYKQVFDRQKKKBANCNFSM4M3GKNLA.

zhangliutong commented 4 years ago

Thank you so much. Do you use the mingboggle 101 in MNI152 space and LPB40 in delineation space?