ComparisonMetrics
=================

Segmentation comparison metrics.

Provides Dice coefficient, Hausdorff distance, and volume measurements
for comparing trial segmentations against gold standards.

Classes
-------

.. py:class:: SegmentationMetrics

   Container for segmentation comparison metrics.

   **Methods:**

   .. py:method:: jaccard()

      Jaccard index (IoU) from Dice coefficient.

   .. py:method:: to_dict()

      Convert to dictionary for JSON serialization.


.. py:class:: ComparisonMetrics

   Calculate segmentation comparison metrics.
   
   Computes Dice, Hausdorff distance, volume, and confusion matrix
   metrics for comparing trial segmentations to gold standards.

   **Methods:**

   .. py:method:: __init__()

      Initialize metrics calculator.

   .. py:method:: compute()

      Compute all comparison metrics.

   .. py:method:: compute_slice_metrics()

      Compute per-slice metrics along an axis.

   .. py:method:: format_summary()

      Format metrics as human-readable summary.

   .. py:method:: last_metrics()

      Get last computed metrics.


Functions
---------

.. py:function:: jaccard()

   Jaccard index (IoU) from Dice coefficient.

.. py:function:: to_dict()

   Convert to dictionary for JSON serialization.

.. py:function:: compute()

   Compute all comparison metrics.
   
   Args:
       trial_mask: Binary trial segmentation mask (3D or 2D).
       gold_mask: Binary gold standard mask (same shape as trial).
       spacing: Voxel spacing in mm (x, y, z). If None, volumes
                and Hausdorff are computed in voxels.
   
   Returns:
       SegmentationMetrics with all computed values.

.. py:function:: compute_slice_metrics()

   Compute per-slice metrics along an axis.
   
   Args:
       trial_mask: 3D binary trial mask.
       gold_mask: 3D binary gold mask.
       axis: Axis to slice along (0=K, 1=J, 2=I for IJK order).
   
   Returns:
       List of dicts with slice index and Dice coefficient.

.. py:function:: format_summary()

   Format metrics as human-readable summary.
   
   Args:
       metrics: Metrics to format. Uses last computed if None.
   
   Returns:
       Formatted string summary.

.. py:function:: last_metrics()

   Get last computed metrics.

.. py:function:: compute_metrics_from_nodes()

   Compute metrics from Slicer segmentation nodes.
   
   Args:
       trial_node: Trial vtkMRMLSegmentationNode.
       gold_node: Gold standard vtkMRMLSegmentationNode.
       reference_volume: Optional reference volume for geometry.
   
   Returns:
       SegmentationMetrics or None if computation fails.

.. py:function:: export_seg_to_array()