Practical Recommendations for dMRI Data Acquisitions

General Recommendations

  • Do not push the resolution. Your images should still have signal above the noise floor in fiber bundles that are parallel to your diffusion gradients.
  • Always acquire isotropic voxel size. For standard MRI scanners with generic gradients, 2mm isotropic resolution is common these days.
  • Do not use very high in-plane acceleration or multi-band acceleration factors. Make sure the acceleration factors you use do not introduce additional imaging artifacts.
  • The diffusion gradients should sample the whole sphere instead of half-sphere to balance the effects of imaging gradients.
  • For brain imaging, try to use a square in-plane field-of-view.
  • The FoV should cover the entire object of interest. Additionally, have a couple of empty-slices before and after (such as inferior and superior in axial acquisitions) the object. This will improve the registration performance.
  • Acquire a fat-suppressed T2W TSE anatomical image for pre-processing.  TORTOISE makes use of this.
  • 1 b=0 s/mm2 image should be acquired for every 6-10 DWI.
  • We STRONGLY recommend to acquire your data with blip-up blip-down acquisitions.


2-way Blip-Up Blip-Down Acquisition Recommendations

  • We STRONGLY recommend to acquire full acquisitions (not just b=0 images) in both the up phase-encoding and down phase-encoding directions
  • For studies that are sensitive to gradient direction sampling resolution,  the up and down data should include different but complimentary diffusion gradient directions.
  • For studies where the SNR of DWIs is important, the up and down data should use identical gradient directions. For this case, TORTOISE will combine the two datasets into a single one with higher SNR in DWIs.

4-way Blip-Up Blip-Down Acquisition Recommendations

  • 4-way phase-encoding refers to acquiring the data using Anterior-Posterior (AP), Posterior-Anterior (PA), Right-Left (RL), Left-Right (LR) phase-encoding directions all together.
  • Diffusion gradient to phase-encoding distribution:
  1. Determine the number of volumes that you can acquire given the scan time. Generate the optimum gradient set for this number of volumes.
  2. For studies that are sensitive to gradient direction sampling resolution,  distribute these gradient/bvals  evenly to the four phase-encoding directions, i.e. each phase-encoding direction should contain the same number of b=0 images and DWIs but with different directions.
  3. For studies where the SNR of DWIs is important,  partition the optimum gradient set into two:   Set1 for AP and PA phase-encoding,  Set 2 for RL and LR encoding.
  4. Concatenate the processed AP/PA and RL/LR datasets into a single, final one to be used in analysis.
  5. Example diffusion sets are provided in this page.


  • For all 4 phase-encoding series,  set TE/TR to minimum achieveable.  Note these values. Compute the final TE (or TR) as TE_{final} = max( TE_{AP}, TE_{PA}, TE_{RL}, TE_{LR} ). Manually set all 4 TEs to this value. Repeat for TR.


Siemens Specific Recommendations

  • Siemens scanner softwares have a bug that change the prescribed phase-encoding directions, when loading an existing acquisition protocol.
  • Make sure that the phase-encoding directions are correct after loading protocols.


GE Specific Recommendations

  • Diffusion gradients for GE are prescribed as read/phase/slice instead of x/y/z. Therefore the 3d gradient vector (g1, g2, g3) with AP phase encoding is not identical to (g1, g2, g3) with RL phase-encoding.  It actually would be (g2, -g1, g3).  Consider this while generating your final gradient sets.