Distortion dominates fibre tracking of the optic chiasm – an evaluation of ultra-high angular resolution compared to low-distortion diffusion MRI on a Compact 3T
Thomas Welton1, Matthew Lyon1, Jerome J Maller1,2, Myung-Ho In3, Ek-Tsoon Tan3, Matt A Bernstein4, Erin M Gray4, Yunhong Shu4, John Huston4, and Stuart M Grieve1,5

1Sydney Translational Imaging Laboratory, Heart Research Institute, University of Sydney, Sydney, Australia, 2GE Healthcare, Richmond, Melbourne, Victoria, Australia, 3GE Global Research, Niskayuna, NY, United States, 4Department of Radiology, Mayo Clinic, Rochester, MN, United States, 5Department of Radiology, Royal Prince Alfred Hospital, Sydney, Australia


We evaluated the impact of angular resolution and spatial distortion on crossing-fibre tracking accuracy at the optic chiasm using diffusion MRI data from a Compact 3T scanner with high-performance gradients. Contralateral tracking via the chiasm was quantified in acquisitions optimised for q-space resolution or low distortion and compared to the known true rate of decussation. We found that, for chiasmal tracking, minimising the effects of geometric distortion may provide better value than maximising spatial or angular resolution beyond 140 directions. An ideal future diffusion MRI protocol will combine these features for more optimal tracking performance.


The inability to resolve crossing fibres is a fundamental limitation of diffusion imaging. It is also known that geometric distortion has a negative impact on diffusion data through voxel compression and displacement. The optic chiasm is the quintessential example of a location where there is both a high density of crossing fibres and high spatial distortion due to the adjacent paranasal sinuses. The anatomically-true proportion of decussating fibres is known from microscopy and tracer studies (53-58%1,2), making the optic chiasm a convenient reference for evaluation of tractography. Recent advances in MRI hardware and acquisition techniques may enable improved tracking performance. Here we sought to investigate the trade-off between angular resolution and distortion in resolving crossing fibres in the chiasm. We hypothesised that high angular resolution diffusion MRI acquired on the latest hardware would translate into improved fibre tracking of the optic chiasm but that these improvements may be limited by the impact of geometric distortion at this location.


One healthy adult subject was imaged under an IRB-approved protocol using a Compact 3T MRI scanner (peak gradient amplitude 80 mT/m, slew rate 700 T/m/s)3-5. To account for additional concomitant fields arising from the asymmetric transverse gradients, frequency shifting6 and gradient pre-emphasis7 was applied. High-order gradient non-linearity correction with even-order terms was applied8. Three diffusion acquisitions were compared:

  • “High angular resolution”: 1.2 mm3, TE=58.6 ms, TR=6000 ms, FA=90°, 750 directions; 3 shells at b=700 (134), 1000 (214) and 2800 (402) mm/s2 plus 42 b=0 volumes, multiband factor=3, in-plane acceleration factor=2, ~80 minutes. This dataset was down-sampled to 33, 64, 140, 280, 420, 560 and 700 directions while retaining uniformly-distributed gradient directions and equal proportions of b-values in each shell.
  • “Low distortion” (MUSE9): 1.2 mm3, TE=54.5 ms, TR=12500 ms, FA=90°, 33 diffusion-weighted volumes at b=1000 mm/s2 plus 1 b=0 volume, in-plane acceleration factor=2, ~15 minutes.
  • “Zero distortion” (DIADEM10,11): 1.5 mm3, signal TE=46.7 ms, navigator TE=64.7 ms TR=8709 ms, FA=90°, 6 diffusion-weighted volumes at b=1000 mm/s2 plus 1 b=0 volume, in-plane acceleration factor=4, ~8 minutes.

Each dataset was denoised, corrected for susceptibility, eddy-currents and motion using TOPUP and eddy_cuda12. A white-matter response function was generated using the Dhollander method and fibre orientation distributions created using constrained spherical deconvolution13. Probabilistic tractography was performed using manually-placed seeds in the optic nerves and tracts. Percentages of tracks crossing to the contralateral hemisphere (average of left and right, nerve-tract and tract-nerve) and reaching the lateral geniculate nuclei (LGN) were measured and compared across acquisitions and down-sampled datasets.


The rate of crossing fibres in all three datasets was underestimated compared to that reported by histological studies by between 1.5 and 3-fold. The low-distortion MUSE dataset performed the best, measuring 30.7% of tracks reaching the contralateral hemisphere, followed by high-angular resolution (23.4%) and DIADEM (15.0%; Figure 1). Increasing angular resolution appeared to have little impact beyond 140 directions, with decussation rates between 22-24% across 140-750 total directions. At 64 and 33 directions, the decussation rate was reduced to 17% and 16%, respectively. Proportions of tracks reaching the LGN were greatest in the high angular resolution dataset (12.2%) compared to MUSE (10.5%) and DIADEM (7.0%), and in the more-dense sampling schemes (r=0.88, p<0.01; ranging from 8.6% at 33 directions to 12.2% at 750 directions).


Our data show that geometric effects dominate fibre tracking performance at the optic chiasm, but that raising angular resolution improves the tracking up to 140 directions. For crossing fibres in the optic chiasm, minimising the effects of geometric distortion may provide better value than maximising the spatial or q-space resolution. For branching fibres in deep white matter, high-angular resolution may provide better value than minimising distortion. Relative to the known proportion of decussating fibres at the chiasm, diffusion datasets underestimated the rate of decussation. This was expected given the inherent difficulty in modelling crossing fibres. The relatively long acquisition time of the DIADEM sequence currently limits angular resolution and, hence, its use in probabilistic tractography applications; however, higher levels of acceleration are possible. The high-performance gradients of a Compact 3T scanner are beneficial for both MUSE and DIADEM acquisitions.


Diffusion acquisitions optimised for reduced geometric distortion greatly enhance tracking of crossing fibres in the optic chiasm – more so than high q-space resolution acquisitions, which perform better when tracking deep white matter bundles. Our results suggest that a MUSE sequence with optimised angular resolution may offer a good performance compromise.


No acknowledgement found.


1. Horton JC: Wilbrand's knee of the primate optic chiasm is an artefact of monocular enucleation. Trans Am Ophthalmol Soc 1997, 95:579-609.

2. Kupfer C, Chumbley L, Downer JC: Quantitative histology of optic nerve, optic tract and lateral geniculate nucleus of man. J Anat 1967, 101(Pt 3):393-401.

3. Foo TKF, Laskaris E, Vermilyea M, Xu M, Thompson P, Conte G, Van Epps C, Immer C, Lee SK, Tan ET et al: Lightweight, compact, and high-performance 3T MR system for imaging the brain and extremities. Magn Reson Med 2018, 80(5):2232-2245.

4. Lee SK, Mathieu JB, Graziani D, Piel J, Budesheim E, Fiveland E, Hardy CJ, Tan ET, Amm B, Foo TK et al: Peripheral nerve stimulation characteristics of an asymmetric head-only gradient coil compatible with a high-channel-count receiver array. Magn Reson Med 2016, 76(6):1939-1950.

5. Weavers PT, Shu Y, Tao S, Huston J, 3rd, Lee SK, Graziani D, Mathieu JB, Trzasko JD, Foo TK, Bernstein MA: Technical Note: Compact three-tesla magnetic resonance imager with high-performance gradients passes ACR image quality and acoustic noise tests. Med Phys 2016, 43(3):1259-1264.

6. Weavers PT, Tao S, Trzasko JD, Frigo LM, Shu Y, Frick MA, Lee SK, Foo TK, Bernstein MA: B0 concomitant field compensation for MRI systems employing asymmetric transverse gradient coils. Magn Reson Med 2018, 79(3):1538-1544.

7. Tao S, Weavers PT, Trzasko JD, Shu Y, Huston J, 3rd, Lee SK, Frigo LM, Bernstein MA: Gradient pre-emphasis to counteract first-order concomitant fields on asymmetric MRI gradient systems. Magn Reson Med 2017, 77(6):2250-2262.

8. Tao S, Trzasko JD, Gunter JL, Weavers PT, Shu Y, Huston J, Lee SK, Tan ET, Bernstein MA: Gradient nonlinearity calibration and correction for a compact, asymmetric magnetic resonance imaging gradient system. Phys Med Biol 2017, 62(2):N18-N31.

9. Chen NK, Guidon A, Chang HC, Song AW: A robust multi-shot scan strategy for high-resolution diffusion weighted MRI enabled by multiplexed sensitivity-encoding (MUSE). Neuroimage 2013, 72:41-47.

10. In MH, Posnansky O, Speck O: High-resolution distortion-free diffusion imaging using hybrid spin-warp and echo-planar PSF-encoding approach. Neuroimage 2017, 148:20-30.

11. In MH, Tan ET, Trzasko JD, Shu Y, Tao S, Gray EM, Huston J, Bernstein MA: Distortion-Free, High-Resolution Diffusion Imaging in a Clinically-Feasible Scan Time on a Compact 3T MRI with High-Performance Gradients. In: ISMRM; Paris. 2018: P.1203.

12. Andersson JLR, Sotiropoulos SN: An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. Neuroimage 2016, 125:1063-1078.

13. Jeurissen B, Tournier JD, Dhollander T, Connelly A, Sijbers J: Multi-tissue constrained spherical deconvolution for improved analysis of multi-shell diffusion MRI data. Neuroimage 2014, 103:411-426.


Figure 1. (A) Illustrative tractograms from each dataset showing the optic chiasm with directionally-encoded colour: A1 (left) High angular resolution, A2 (middle) MUSE, A3 (right) DIADEM. (B) Proportions of tracks reaching the contralateral hemisphere compared across datasets. (C) Proportions of tracks reaching the contralateral hemisphere for each subsample of the high-angular resolution dataset. (D) Proportions of tracks reaching the lateral geniculate nucleus for each subsample of the high-angular resolution dataset, with the 750-direction sequence and subsamples in blue, the DIADEM in green and MUSE in red.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)