fix: for .PAR headers, avoid the assumption of 4D shape in get_bvals_vecs

For example, post-processed ADC maps will be marked as diffusion scans but only have 3D
data. The b-values and vectors from the original scan are not available in the header in
this case.

Author: grlee77

nibabel/parrec.py

@@ -784,7 +784,13 @@ def get_bvals_bvecs(self):
         if self.general_info['diffusion'] == 0:
             return None, None
         reorder = self.get_sorted_slice_indices()
-        n_slices, n_vols = self.get_data_shape()[-2:]
+        if len(self.get_data_shape()) == 3:
+            # Any original diffusion scans will have >=2 volumes. However, a
+            # single dynamic is possible for a post-processed difusion volume
+            # such as an ADC map. The b-values are unavailble in this case.
+            return None, None
+        else:
+            n_slices, n_vols = self.get_data_shape()[-2:]
         bvals = self.image_defs['diffusion_b_factor'][reorder].reshape(
             (n_slices, n_vols), order='F')
         # All bvals within volume should be the same