Merge pull request #40 from reghbali/reghbali_test_qunatification_tas…

…k_run

Reghbali test qunatification task run

.github/workflows/integration-test.yml

@@ -63,6 +63,10 @@ jobs:
               cp c3d*/bin/c3d $BIN/c3d
           fi
 
+      - name: Download models
+        run: |
+          pyalfe download models
+
       - name: Run tests
         run: |
           pytest tests/integration

pyalfe/models/__init__.py

@@ -1,7 +1,7 @@
 import os
 from pathlib import Path
 
-models_url = 'https://ucsf.box.com/shared/static/' 'a93ruk3m26mso38jm8gk3qs5iod0h90h.gz'
+models_url = 'https://ucsf.box.com/shared/static/a93ruk3m26mso38jm8gk3qs5iod0h90h.gz'
 
 MODELS_PATH = Path(os.path.expanduser(os.path.join('~', '.cache', 'pyalfe')))
 

pyalfe/tasks/quantification.py

@@ -27,7 +27,7 @@ class Quantification(Task):
         All the modalities for processed by the pipeline.
     modalities_target: list[Modality]
         Target modalities that are used to defince lesions.
-    dominant_tissue: bool
+    dominant_tissue: str
         The dominant tissue where the tumor or lesion is expected to be
         located at. THe options are `white_matter`, `gray_matter`, `auto`.
         If `auto` is chosen, then the dominant tissue is chosen to be the tissue
@@ -41,7 +41,7 @@ def __init__(
         pipeline_dir: PipelineDataDir,
         modalities_all: list[Modality],
         modalities_target: list[Modality],
-        dominant_tissue: bool = None,
+        dominant_tissue: str = None,
     ):
         self.modalities_all = modalities_all
         self.modalities_target = modalities_target
@@ -81,26 +81,33 @@ def load_modality_images(self, accession, target):
 
             if os.path.exists(modality_image_to_target_file):
                 modality_images[modality], _ = self.load(modality_image_to_target_file)
-            else:
-                modality_images[modality] = None
         return modality_images
 
-    def load_template_images(self, accession, target):
+    def load_template_images(self, accession, modality):
         target_images = {}
 
         for roi_key, roi_properties in roi_dict.items():
 
             roi_sub_dir = roi_properties['sub_dir']
 
             if roi_properties['type'] == 'template':
-                template_image_to_target_file = self.pipeline_dir.get_output_image(
-                    accession,
-                    modality=target,
-                    image_type=roi_key,
-                    resampling_target=target,
-                    resampling_origin=Modality.T1,
-                    sub_dir_name=roi_sub_dir,
-                )
+                if modality == Modality.T1:
+                    template_image_to_target_file = self.pipeline_dir.get_output_image(
+                        accession,
+                        modality=modality,
+                        resampling_target=modality,
+                        resampling_origin=roi_key,
+                        sub_dir_name=roi_sub_dir,
+                    )
+                else:
+                    template_image_to_target_file = self.pipeline_dir.get_output_image(
+                        accession,
+                        modality=modality,
+                        image_type=roi_key,
+                        resampling_target=modality,
+                        resampling_origin=Modality.T1,
+                        sub_dir_name=roi_sub_dir,
+                    )
 
                 if os.path.exists(template_image_to_target_file):
                     target_images[roi_key], _ = self.load(template_image_to_target_file)
@@ -112,7 +119,6 @@ def load_template_images(self, accession, target):
                     self.logger.info(
                         'missing template: ' + template_image_to_target_file
                     )
-                    target_images[roi_key] = None
         return target_images
 
     def get_radiomics(self, skullstripped_file, lesion_seg_file):
@@ -241,6 +247,11 @@ def get_lesion_stats(
             stats['enhancement'] = np.mean(t1post_image[lesion_indices]) / np.mean(
                 t1_image[lesion_indices]
             )
+        else:
+            self.logger.info(
+                'Cannot quantify avg enhancement over the leasion because T1'
+                ' or T1Post are not registered to target modality'
+            )
 
         if ventricles_distance is not None:
             stats['average_dist_to_ventricles_(voxels)'] = np.mean(
@@ -284,32 +295,38 @@ def run(self, accession):
             accession=accession, modality=Modality.T1, image_type='skullstripping_mask'
         )
 
-        brain_mask, voxel_volume = self.load(brain_mask_file)
+        if not os.path.exists(brain_mask_file):
+            self.logger.info(
+                'T1 skullstripping mask is missing.'
+                'Skipping brain volume quantification.'
+            )
+        else:
+            brain_mask, voxel_volume = self.load(brain_mask_file)
 
-        tissue_seg_file = self.pipeline_dir.get_output_image(
-            accession=accession, modality=Modality.T1, image_type='tissue_seg'
-        )
+            tissue_seg_file = self.pipeline_dir.get_output_image(
+                accession=accession, modality=Modality.T1, image_type='tissue_seg'
+            )
 
-        if not os.path.exists(tissue_seg_file):
-            tissue_seg = None
-        else:
-            tissue_seg, _ = self.load(tissue_seg_file)
+            if not os.path.exists(tissue_seg_file):
+                tissue_seg = None
+            else:
+                tissue_seg, _ = self.load(tissue_seg_file)
 
-        template_images = self.load_template_images(accession, Modality.T1)
+            template_images = self.load_template_images(accession, Modality.T1)
 
-        ventricles_seg_file = self.pipeline_dir.get_output_image(
-            accession=accession, modality=Modality.T1, image_type='VentriclesSeg'
-        )
+            ventricles_seg_file = self.pipeline_dir.get_output_image(
+                accession=accession, modality=Modality.T1, image_type='VentriclesSeg'
+            )
 
-        if not os.path.exists(ventricles_seg_file):
-            ventricles_seg = None
-        else:
-            ventricles_seg, _ = self.load(ventricles_seg_file)
+            if not os.path.exists(ventricles_seg_file):
+                ventricles_seg = None
+            else:
+                ventricles_seg, _ = self.load(ventricles_seg_file)
 
-        volume_stats = self.get_brain_volume_stats(
-            brain_mask, tissue_seg, ventricles_seg, voxel_volume
-        )
-        pd.Series(volume_stats).to_csv(volumetric_quantification_file)
+            volume_stats = self.get_brain_volume_stats(
+                brain_mask, tissue_seg, ventricles_seg, template_images, voxel_volume
+            )
+            pd.Series(volume_stats).to_csv(volumetric_quantification_file)
 
         for target in self.modalities_target:
             summary_quantification_file = self.pipeline_dir.get_quantification_file(
@@ -356,7 +373,6 @@ def run(self, accession):
             )
 
             if not os.path.exists(tissue_seg_file):
-                print('no tissue seg', tissue_seg_file)
                 self.logger.info(
                     f'Tissue seg file for {target} is missing.'
                     f'Skipping quantification for {target}.'

tests/unit/test_image_registration.py

@@ -93,7 +93,7 @@ def test_register_affine(self):
             np.testing.assert_array_less(
                 np.linalg.norm(get_nifti_data(output) - fixed_data)
                 / np.linalg.norm(get_nifti_data(output) - moving_data),
-                0.45,
+                0.47,
             )
 
             moving2_path = self.get_image_path('moving2.nii.gz')