point set conversion

README.md

@@ -6,32 +6,223 @@
 
 Convert between [itk](https://itk.org) and [napari](https://napari.org) data structures.
 
-Installation
-------------
+## Installation
 
-```
+```sh
 pip install itk-napari-conversion
 ```
 
-Usage
------
+## Usage
 
-Convert an `itk.Image` to an `napari.layers.Image`:
+### Image Conversion
 
-```
+#### Convert ITK Image to napari Image Layer
+
+Convert an `itk.Image` to a `napari.layers.Image`:
+
+```python
 from itk_napari_conversion import image_layer_from_image
 
 image_layer = image_layer_from_image(image)
 ```
 
-Convert to an `napari.layers.Image` to an `itk.Image`:
+**Features:**
+- Automatically detects and handles RGB/RGBA images
+- Preserves image metadata (spacing, origin, direction, custom metadata)
+- Converts ITK's physical space information to napari's layer transformations:
+  - `spacing` → `scale`
+  - `origin` → `translate`
+  - `direction` → `rotate`
+
+**Example:**
+```python
+import itk
+import napari
+from itk_napari_conversion import image_layer_from_image
+
+# Read an image
+image = itk.imread('path/to/image.nrrd')
+
+# Add custom metadata
+image['units'] = 'mm'
+image['patient_id'] = '12345'
+
+# Convert to napari layer
+image_layer = image_layer_from_image(image)
+
+# The layer will have:
+# - image_layer.scale = image spacing (in reverse order for NumPy)
+# - image_layer.translate = image origin (in reverse order for NumPy)
+# - image_layer.rotate = image direction matrix (in reverse order for NumPy)
+# - image_layer.metadata = all custom metadata
 ```
+
+#### Convert napari Image Layer to ITK Image
+
+Convert a `napari.layers.Image` to an `itk.Image`:
+
+```python
 from itk_napari_conversion import image_from_image_layer
 
 image = image_from_image_layer(image_layer)
 ```
 
-Hacking
+**Features:**
+- Automatically handles RGB/RGBA layers
+- Converts napari layer transformations back to ITK physical space:
+  - `scale` → `spacing`
+  - `translate` → `origin`
+  - `rotate` → `direction`
+- Preserves all metadata from the layer
+
+**Example:**
+```python
+import numpy as np
+import napari
+from itk_napari_conversion import image_from_image_layer
+
+# Create a napari image layer with transformations
+viewer = napari.Viewer()
+data = np.random.rand(100, 100, 100)
+layer = viewer.add_image(
+    data,
+    scale=[2.0, 1.5, 1.5],  # anisotropic spacing
+    translate=[10.0, 20.0, 30.0],
+    metadata={'description': 'My volume'}
+)
+
+# Convert to ITK
+image = image_from_image_layer(layer)
+
+# The ITK image will have:
+# - spacing: coordinates in ITK order, so reversed from napari `scale`: [1.5, 1.5, 2.0]
+# - origin: coordinates in ITK order, so reversed from napari `translate`: [30.0, 20.0, 10.0]
+# - direction: from napari `rotate`. In this case unspecified, so just the identity matrix.
+# - metadata dictionary with custom fields
+```
+
+### Point Set Conversion
+
+#### Convert ITK PointSet to napari Points Layer
+
+Convert an `itk.PointSet` to a `napari.layers.Points`:
+
+```python
+from itk_napari_conversion import points_layer_from_point_set
+
+points_layer = points_layer_from_point_set(point_set)
+```
+
+**Features:**
+- Extracts point coordinates from ITK PointSet using `itk.array_from_vector_container()`
+- Converts point data (if present) to napari features dictionary
+  - Uses the first component as the 'feature' key
+- Returns a `napari.layers.Points` object
+
+**Example:**
+```python
+import itk
+import numpy as np
+from itk_napari_conversion import points_layer_from_point_set
+
+# Create ITK PointSet
+PointSetType = itk.PointSet[itk.F, 3]
+point_set = PointSetType.New()
+
+# Add points
+points_data = np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]], dtype=np.float32)
+points = itk.vector_container_from_array(points_data.flatten())
+point_set.SetPoints(points)
+
+# Add point data (features)
+feature_data = np.array([10.0, 20.0], dtype=np.float32)
+point_data = itk.vector_container_from_array(feature_data)
+point_set.SetPointData(point_data)
+
+# Convert to napari
+points_layer = points_layer_from_point_set(point_set)
+# points_layer.features['feature'] will contain [10.0, 20.0]
+```
+
+#### Convert napari Points Layer to ITK PointSet
+
+Convert a `napari.layers.Points` to an `itk.PointSet`:
+
+```python
+from itk_napari_conversion import point_set_from_points_layer
+
+point_set = point_set_from_points_layer(points_layer)
+```
+
+**Features:**
+- Applies napari transformations (scale, rotate, translate) to point coordinates before conversion
+- Extracts points from napari layer data
+- Converts the first feature column (if present) to ITK point data
+- Returns an `itk.PointSet` object with dimension determined from the data
+
+**Example:**
+```python
+import napari
+import numpy as np
+from itk_napari_conversion import point_set_from_points_layer
+
+# Create napari Points layer with transformations
+data = np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]])
+features = {'intensity': np.array([10.0, 20.0])}
+scale = np.array([2.0, 2.0, 2.0])
+translate = np.array([100.0, 200.0, 300.0])
+
+# Optional: add rotation (90 degrees around z-axis)
+angle = np.radians(90)
+rotate = np.array([
+    [np.cos(angle), -np.sin(angle), 0],
+    [np.sin(angle), np.cos(angle), 0],
+    [0, 0, 1]
+])
+
+points_layer = napari.layers.Points(
+    data,
+    features=features,
+    scale=scale,
+    rotate=rotate,
+    translate=translate
+)
+
+# Convert to ITK PointSet
+point_set = point_set_from_points_layer(points_layer)
+
+# The points in the ITK PointSet will be in world coordinates:
+# ((data * scale) @ rotate.T) + translate
+# Point data will be stored from the 'intensity' feature
+```
+
+### Transformation Handling
+
+#### Images
+- **ITK → napari**: Physical space metadata (spacing, origin, direction) is converted to napari layer transformations
+- **napari → ITK**: Layer transformations are converted back to ITK physical space metadata
+
+#### Points
+- **ITK → napari**: Points are copied as-is without transformations (identity transform assumed)
+- **napari → ITK**: Layer transformations (scale, rotate, translate) are **applied to points** to convert them to world coordinates
+  - Transformations are applied in order: scale → rotate → translate
+
+### Notes
+
+**Images:**
+- Supports 2D, 3D, and multi-dimensional images
+- RGB and RGBA images are automatically detected and handled
+- Axis order is automatically reversed between ITK (x, y, z) and NumPy/napari (z, y, x)
+- Metadata is preserved bidirectionally
+
+**Points:**
+- Point data in ITK is stored as float32
+- Only the first feature column from napari is used for ITK point data
+- Empty point sets are handled gracefully
+- Dimension is automatically determined from the point data shape
+- Metadata conversion is not currently supported for point sets
+
+## Hacking
 -------
 
 Contributions are welcome!

itk_napari_conversion.py

@@ -1,8 +1,13 @@
 """Convert between itk and napari data structures."""
 
-__version__ = "0.3.1"
+__version__ = "0.4.0"
 
-__all__ = ["image_layer_from_image", "image_from_image_layer"]
+__all__ = [
+    "image_layer_from_image",
+    "image_from_image_layer",
+    "points_layer_from_point_set",
+    "point_set_from_points_layer",
+]
 
 import napari
 import itk
@@ -55,3 +60,90 @@ def image_from_image_layer(image_layer):
         image["direction"] = np.ascontiguousarray(np.transpose(image_layer.rotate)).astype(np.float64)
 
     return image
+
+
+def points_layer_from_point_set(point_set):
+    """Convert an itk.PointSet to a napari.layers.Points."""
+    # Get points as numpy array
+    number_of_points = point_set.GetNumberOfPoints()
+
+    if number_of_points == 0:
+        data = np.array([]).reshape(0, 3)  # Default to 3D empty array
+    else:
+        points_array = itk.array_from_vector_container(point_set.GetPoints())
+        data = points_array
+
+    # Get point data (features) if available
+    point_data = point_set.GetPointData()
+    features = None
+    if point_data.Size() > 0:
+        point_data_array = itk.array_from_vector_container(point_data)
+        # Use first component as feature if it's a scalar
+        if point_data_array.ndim == 1:
+            features = {'feature': point_data_array}
+        else:
+            # If multi-component, use first component
+            features = {'feature': point_data_array[:, 0] if point_data_array.shape[1] > 0 else point_data_array[:, 0:1].flatten()}
+
+    points_layer = napari.layers.Points(
+        data,
+        features=features,
+    )
+    return points_layer
+
+
+def point_set_from_points_layer(points_layer):
+    """Convert a napari.layers.Points to an itk.PointSet."""
+    # Apply transformations (rotate, scale, translate) to points
+    data = points_layer.data.copy()  # Make a copy to avoid modifying original
+
+    # Napari always has an affine that includes scale and translate
+    # We need to apply it to get world coordinates
+    # Apply transformations in order: scale, rotate, translate
+    if points_layer.scale is not None:
+        data = data * points_layer.scale
+    if points_layer.rotate is not None:
+        # Apply rotation matrix to each point
+        rotate_matrix = np.asarray(points_layer.rotate)
+        data = data @ rotate_matrix.T
+    if points_layer.translate is not None:
+        data = data + points_layer.translate
+
+    # Determine dimension from data
+    if len(data) == 0:
+        dimension = 3  # Default to 3D
+    else:
+        dimension = data.shape[1]
+
+    # Create ITK PointSet
+    # Use float pixel type for point data by default
+    PointSetType = itk.PointSet[itk.F, dimension]
+    point_set = PointSetType.New()
+
+    # Set points
+    if len(data) > 0:
+        points = itk.vector_container_from_array(data.astype(np.float32).flatten())
+        point_set.SetPoints(points)
+
+    # Set point data from features if available
+    if points_layer.features is not None and len(points_layer.features) > 0:
+        feature_keys = list(points_layer.features.keys())
+        if len(feature_keys) > 0:
+            # Use the first feature column as point data
+            feature_name = feature_keys[0]
+
+            # Handle both dict and DataFrame
+            if hasattr(points_layer.features, 'iloc'):
+                # It's a pandas DataFrame
+                feature_data = points_layer.features[feature_name].values
+            else:
+                # It's a dict
+                feature_data = points_layer.features[feature_name]
+
+            if isinstance(feature_data, (list, np.ndarray)):
+                feature_array = np.array(feature_data).astype(np.float32)
+                if len(feature_array) > 0:
+                    point_data = itk.vector_container_from_array(feature_array)
+                    point_set.SetPointData(point_data)
+
+    return point_set