Add RandomBiasFieldDenoise and fix Colin27 exampleFix/colin27 example

src/torchio/transforms/__init__.py

@@ -8,6 +8,7 @@
 from .augmentation.intensity import Motion
 from .augmentation.intensity import Noise
 from .augmentation.intensity import RandomBiasField
+from .augmentation.intensity import RandomBiasFieldDenoise
 from .augmentation.intensity import RandomBlur
 from .augmentation.intensity import RandomGamma
 from .augmentation.intensity import RandomGhosting
@@ -120,4 +121,5 @@
     'RemoveLabels',
     'SequentialLabels',
     'KeepLargestComponent',
+    'RandomBiasFieldDenoise',
 ]

src/torchio/transforms/augmentation/intensity/__init__.py

@@ -1,5 +1,6 @@
 from .random_bias_field import BiasField
 from .random_bias_field import RandomBiasField
+from .random_biasfield_denoise import RandomBiasFieldDenoise
 from .random_blur import Blur
 from .random_blur import RandomBlur
 from .random_gamma import Gamma
@@ -36,4 +37,5 @@
     'BiasField',
     'RandomLabelsToImage',
     'LabelsToImage',
+    'RandomBiasFieldDenoise',
 ]

src/torchio/transforms/augmentation/intensity/random_biasfield_denoise.py

@@ -0,0 +1,34 @@
+from torchio.data.subject import Subject
+
+from ...transform import Transform
+
+
+class RandomBiasFieldDenoise(Transform):
+    """
+    Simple placeholder transform that simulates denoising after bias field
+    correction by blending voxel intensities toward the mean value.
+
+    Parameters:
+        noise_reduction_factor (float): Strength of denoising (0-1).
+    """
+
+    def __init__(self, noise_reduction_factor: float = 0.1, **kwargs):
+        super().__init__(**kwargs)
+        self.noise_reduction_factor = noise_reduction_factor
+
+    def apply_transform(self, subject: Subject) -> Subject:
+        for _, image in subject.get_images_dict(intensity_only=True).items():
+            tensor = image.data.float()
+
+            # Basic denoising by shifting toward mean intensity
+            mean_val = tensor.mean()
+            tensor = (tensor * (1 - self.noise_reduction_factor)) + (
+                mean_val * self.noise_reduction_factor
+            )
+
+            image.set_data(tensor)
+
+        return subject
+
+    def __repr__(self):
+        return f'{self.__class__.__name__}(noise_reduction_factor={self.noise_reduction_factor})'

visualize_denoise.py

@@ -0,0 +1,34 @@
+import matplotlib.pyplot as plt
+
+import torchio as tio
+
+subject = tio.datasets.Colin27()
+
+# Get full 3D volume
+image_3d = subject.t1.data.squeeze().numpy()
+
+# Pick center slice (axial)
+slice_idx = image_3d.shape[2] // 2
+image = image_3d[:, :, slice_idx]
+
+# ---- Apply custom transform ----
+transform = tio.RandomBiasFieldDenoise(noise_reduction_factor=0.3)
+denoised_subject = transform(subject)
+
+# Extract denoised slice
+denoised_3d = denoised_subject.t1.data.squeeze().numpy()
+denoised_image = denoised_3d[:, :, slice_idx]
+
+# ---- Plot ----
+fig, axes = plt.subplots(1, 2, figsize=(10, 5))
+
+axes[0].imshow(image, cmap='gray')
+axes[0].set_title('Original')
+axes[0].axis('off')
+
+axes[1].imshow(denoised_image, cmap='gray')
+axes[1].set_title('After Denoise')
+axes[1].axis('off')
+
+plt.tight_layout()
+plt.show()