fixing indentation and formatting

Signed-off-by: Vijay Vignesh Prasad Rao <vijayvigneshp02@gmail.com>

Author: VijayVignesh1

monai/metrics/meandice.py

@@ -13,20 +13,14 @@
 
 import numpy as np
 import torch
-import torch.nn.functional as F
-
 from scipy.ndimage import distance_transform_edt, generate_binary_structure
 from scipy.ndimage import label as sn_label
 
 from monai.metrics.utils import do_metric_reduction
 from monai.utils import MetricReduction, deprecated_arg
-from monai.utils.module import optional_import
 
 from .metric import CumulativeIterationMetric
 
-distance_transform_edt, has_ndimage = optional_import("scipy.ndimage", name="distance_transform_edt")
-sn_label, _ = optional_import("scipy.ndimage", name="label")
-
 __all__ = ["DiceMetric", "compute_dice", "DiceHelper"]
 
 
@@ -309,15 +303,11 @@ def compute_voronoi_regions_fast(self, labels, connectivity=26, sampling=None):
         Returns the ID of the nearest component for each voxel.
 
         Args:
-            labels (np.ndarray | torch.Tensor): Label map where values > 0 are seeds.
-            connectivity (int): 6, 18, or 26 for 3D connectivity. Defaults to 26.
-            sampling (tuple[float, ...] | None): Voxel spacing for anisotropic distances.
-
-        Returns:
-            torch.Tensor: Voronoi region IDs (int32) on CPU.
+            labels: input label map as a numpy array, where values > 0 are considered seeds for connected components.
+            connectivity: 6/18/26 (3D)
+            sampling: voxel spacing for anisotropic distances (scipy.ndimage.distance_transform_edt)
         """
-        if not has_ndimage:
-            raise RuntimeError("scipy.ndimage is required for per_component Dice computation.")
+
         x = np.asarray(labels)
         conn_rank = {6: 1, 18: 2, 26: 3}.get(connectivity, 3)
         structure = generate_binary_structure(rank=3, connectivity=conn_rank)
@@ -332,14 +322,12 @@ def compute_voronoi_regions_fast(self, labels, connectivity=26, sampling=None):
 
     def compute_cc_dice(self, y_pred: torch.Tensor, y: torch.Tensor) -> torch.Tensor:
         """
-        Compute per-component Dice for a single batch item.
+        Compute the dice metric for binary inputs which have only spatial dimensions. This method is called separately
+        for each batch item and for each channel of those items.
 
         Args:
-            y_pred (torch.Tensor): Predictions with shape (1, 2, D, H, W).
-            y (torch.Tensor): Ground truth with shape (1, 2, D, H, W).
-
-        Returns:
-            torch.Tensor: Mean Dice over connected components.
+            y_pred: input predictions with shape HW[D].
+            y: ground truth with shape HW[D].
         """
         data = []
         if y_pred.ndim == y.ndim:
@@ -349,9 +337,7 @@ def compute_cc_dice(self, y_pred: torch.Tensor, y: torch.Tensor) -> torch.Tensor
             y_pred_idx = y_pred
             y_idx = y
         if y_idx[0].sum() == 0:
-            if self.ignore_empty:
-                data.append(torch.tensor(float("nan"), device=y_idx.device))
-            elif y_pred_idx.sum() == 0:
+            if y_pred_idx.sum() == 0:
                 data.append(torch.tensor(1.0, device=y_idx.device))
             else:
                 data.append(torch.tensor(0.0, device=y_idx.device))

tests/metrics/test_compute_meandice.py

@@ -18,7 +18,6 @@
 from parameterized import parameterized
 
 from monai.metrics import DiceHelper, DiceMetric, compute_dice
-from monai.metrics.fid import FIDMetric
 
 _device = "cuda:0" if torch.cuda.is_available() else "cpu"
 # keep background