For single shell DTI experiments dcm2nii porduces two versions of the data set, the full data and one with the last volume deleted (indicated with filename starting with x).
However for multi shell DTI data the calculated isotropic images is produced for each b-value by the philips system and not only for the highest b-value.
These should all be deleted in the "x" marked nii bval an bvec files and not only the last volume that contains the calculated isotropic image for the highest b-value.

These calculated isotropic volumes can be identified by having a non zero b-value and a gradient direction of (0,0,0).

Hello-

  I do not have a Philips system, so I would appreciate a sample dataset that illustrates this behavior. Here is my understanding:

 When Philips saves a DTI dataset it can compute derived images from the dataset, for example an ADC map. Most DTI processing tools only expect raw data (B0 images and directional images), and therefore it is useful to separate these images from the raw dataset. To be clear, for DICOM datasets Philips is clearly violating the DICOM standard: derived data should be saved with a different series number to distinguish it from the raw data.
 
-c