DiceMetric to compute separate organs dice score

[FerseinChris]

Is there a way to use Monai and obtain the dice of each organ in my volume (case of BTCV dataset)? Below is the code used to compute dice.

def validation(epoch_iterator_val): model.eval() with torch.no_grad(): for batch in epoch_iterator_val: val_inputs, val_labels = (batch["image"].cuda(), batch["label"].cuda()) val_outputs = sliding_window_inference(val_inputs, (96, 96, 96), 4, model) val_labels_list = decollate_batch(val_labels) val_labels_convert = [post_label(val_label_tensor) for val_label_tensor in val_labels_list] val_outputs_list = decollate_batch(val_outputs) val_output_convert = [post_pred(val_pred_tensor) for val_pred_tensor in val_outputs_list] dice_metric(y_pred=val_output_convert, y=val_labels_convert) epoch_iterator_val.set_description("Validate (%d / %d Steps)" % (global_step, 10.0)) # noqa: B038 mean_dice_val = dice_metric.aggregate().item() dice_metric.reset() return mean_dice_val

[KumoLiu]

Hi @FerseinChris, you can get separate dice for each channel by setting the reduction to mean_batch.

dice_metric = DiceMetric(include_background=True, reduction="mean_batch", get_not_nans=False)

For how to process your output, you can refer to this tutorial: https://github.com/Project-MONAI/tutorials/blob/main/3d_segmentation/swin_unetr_btcv_segmentation_3d.ipynb
Hope it helps, thanks.

[FerseinChris]

Thanks dear for the fast reply! Based on the example code provided through the link, I wish to know how to get the values for each dice of each organ to segment?

[KumoLiu]

you can get separate dice for each channel by setting the reduction to mean_batch.
dice_metric = DiceMetric(include_background=True, reduction="mean_batch", get_not_nans=False)

[FerseinChris]

I've tried this way you provided and I've encountered an error. I think it should be better you provided a code using mean_batch as reduction. Because the code on the link rather use mean.

…

On Mon, Aug 26, 2024 at 2:20 PM Jorvialom Tshivetta ***@***.***> wrote: Thanks dear for the fast reply! Based on the example code provided through the link, I wish to know how to get the values for each dice of each organ to segment?

[KumoLiu]

Hi @FerseinChris,
You can refer to these tutorials, they all use mean_batch as reduction.
https://github.com/Project-MONAI/tutorials/blob/main/3d_segmentation/swin_unetr_brats21_segmentation_3d.ipynb
https://github.com/Project-MONAI/tutorials/blob/main/competitions/MICCAI/surgtoolloc/classification_files/train.py
https://github.com/Project-MONAI/tutorials/blob/main/acceleration/distributed_training/brats_training_ddp.py

[FerseinChris]

Thanks a lot.

…

On Mon, Aug 26, 2024 at 2:33 PM YunLiu ***@***.***> wrote: Hi @FerseinChris <https://github.com/FerseinChris>, You can refer to these tutorials, they all use mean_batch as reduction. https://github.com/Project-MONAI/tutorials/blob/main/3d_segmentation/swin_unetr_brats21_segmentation_3d.ipynb https://github.com/Project-MONAI/tutorials/blob/main/competitions/MICCAI/surgtoolloc/classification_files/train.py https://github.com/Project-MONAI/tutorials/blob/main/acceleration/distributed_training/brats_training_ddp.py — Reply to this email directly, view it on GitHub <#1800 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AQHEBDKYIPOXJQ6UGU5MHLTZTLD3JAVCNFSM6AAAAABNDDDXA6VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTANBUG43TAMI> . You are receiving this because you were mentioned.Message ID: ***@***.*** com>