Removed duplicate code, extraneous logger line

niCHARTPipelines/CalculateROIVolume.py

@@ -24,7 +24,7 @@ def calc_roi_volumes(mrid, in_img_file, label_indices = []):
 
     ## Get label indices
     if label_indices.shape[0] == 0:
-        logger.warning('Label indices not provided, generating from data')
+        # logger.warning('Label indices not provided, generating from data')
         label_indices = u_ind
 
     label_names = label_indices.astype(str)
@@ -99,3 +99,30 @@ def create_roi_csv(scan_id, in_roi, list_single_roi, map_derived_roi, out_img, o
 
     ## Write out csv
     df_dmuse.to_csv(out_csv, index = False)
+
+###---------calculate ROI volumes-----------
+def extract_roi_masks(in_roi, map_derived_roi, out_pref):
+    '''Create individual roi masks for single and derived rois
+    '''
+    img_ext_type = '.nii.gz'
+
+    ## Read image
+    in_nii = nib.load(in_roi)
+    img_mat = in_nii.get_fdata().astype(int)
+
+    ## Read derived roi map file to a dictionary
+    roi_dict = {}
+    with open(map_derived_roi) as roi_map:
+        reader = csv.reader(roi_map, delimiter=',')
+        for row in reader:
+            key = str(row[0])
+            val = [int(x) for x in row[2:]]
+            roi_dict[key] = val
+
+    # Create an individual roi mask for each roi
+    for i, key in enumerate(roi_dict):
+        print(i)
+        key_vals = roi_dict[key]
+        tmp_mask = np.isin(img_mat, key_vals).astype(int)
+        out_nii = nib.Nifti1Image(tmp_mask, in_nii.affine, in_nii.header)
+        nib.save(out_nii, out_pref + '_' + str(key) + img_ext_type)

niCHARTPipelines/MaskImage.py

@@ -3,47 +3,7 @@
 from scipy import ndimage
 from scipy.ndimage.measurements import label
 
-## Find bounding box for the foreground values in img, with a given padding percentage
-def calc_bbox_with_padding(img, perc_pad = 10):
-
-    img = img.astype('uint8')
-
-    ## Output is the coordinates of the bounding box
-    bcoors = np.zeros([3,2], dtype=int)
-
-    ## Find the largest connected component 
-    ## INFO: In images with very large FOV DLICV may have small isolated regions in
-    ##       boundaries; so we calculate the bounding box based on the brain, not all
-    ##       foreground voxels
-    str_3D = np.array([[[0, 0, 0], [0, 1, 0], [0, 0, 0]],
-                       [[0, 1, 0], [1, 1, 1], [0, 1, 0]],
-                       [[0, 0, 0], [0, 1, 0], [0, 0, 0]]], dtype='uint8')
-    labeled, ncomp = label(img, str_3D)
-    sizes = ndimage.sum(img, labeled, range(ncomp + 1))
-    img_largest_cc = (labeled == np.argmax(sizes)).astype(int)
-
-    ## Find coors in each axis
-    for sel_axis in [0, 1, 2]:
-
-        ## Get axes other than the selected
-        other_axes = [0, 1, 2]
-        other_axes.remove(sel_axis)
-
-        ## Get img dim in selected axis
-        dim = img_largest_cc.shape[sel_axis]
-
-        ## Find bounding box (index of first and last non-zero slices)
-        nonzero = np.any(img_largest_cc, axis = tuple(other_axes))
-        bbox= np.where(nonzero)[0][[0,-1]]    
-
-        ## Add padding
-        size_pad = int(np.round((bbox[1] - bbox[0]) * perc_pad / 100))
-        b_min = int(np.max([0, bbox[0] - size_pad]))
-        b_max = int(np.min([dim, bbox[1] + size_pad]))
-
-        bcoors[sel_axis, :] = [b_min, b_max]
-
-    return bcoors
+from niCHARTPipelines.CombineMasks import calc_bbox_with_padding
 
 
 ###---------mask image-----------