Giorgia Grisot1,2, Julia Lehman3, Suzanne N Haber3, and Anastasia Yendiki2
1Harvard-MIT Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, 2Athinoula A. Martinos Center for Biomedical Imaging, MGH, Charlestown, MA, United States, 3University of Rochester School of Medicine, Rochester, NY, United States
Synopsis
Determining the optimal diffusion MRI (dMRI)
acquisition scheme for reconstructing a brain network of interest with
tractography is an open problem, and the lack of ground truth on brain
connections makes it challenging to resolve. We use chemical tracing and ex
vivo dMRI in macaques to optimize dMRI acquisition with respect to tractography
accuracy. We present preliminary results illustrating that 1. There is an upper
bound to the angular resolution of dMRI, beyond which tractography accuracy
does not improve, and 2. That this finding likely generalizes to in vivo human
dMRI.Purpose
Determining the optimal diffusion MRI (dMRI)
acquisition scheme for reconstructing a brain network of interest for a
specific study is an open problem. Given the time limitations of in-vivo scans,
it is important to determine which parameters (b-values, number of gradient
directions and number of shells) have the greatest effect on the accuracy of white
matter (WM) pathways reconstructed by tractography. However, the lack of ground
truth on brain circuitry poses a challenge for quantifying the accuracy of dMRI
tractography. Here, we propose to use chemical tracing in macaques for
optimizing dMRI acquisitions with respect to tractography accuracy. Previous
studies that performed dMRI validation using chemical tracing did not compare
acquisition parameters, using instead a single dMRI acquisition. These studies either
used dMRI with low angular resolution1,23 or used tracer and
dMRI data from different animals1,4,5. Here, we collect
ex-vivo dMRI data with 22 shells, spanning b-values from 1600 to 40k, on a
macaque brain that has received a tracer injection in the frontal cortex. This
allows us to investigate how the addition of each shell affects tractography
accuracy. We find that there are diminishing returns beyond a b-value of 25.6k
for reconstructing projections from this injection site. We then show a
preliminary evaluation of in-vivo human dMRI tractography from the homologous
cortical region with different shells.Methods
Macaque
data: An adult macaque was injected with Lucifer yellow in
BA10. Following injection and perfusion, the brain was imaged in a small-bore 4.7T
MRI system with maximum gradient=480mT/m. A dMRI data set was collected using a
2-shot EPI sequence with δ=15ms, Δ=19ms, 514 directions, 0.7mm resolution and bmax=40000s/mm2.
The brain was then sectioned and stained. Axon bundles were manually traced on
the histology data using Neurolucida and spatially aligned to the dMRI data
using our in-house pipeline (Fig.2). We extracted subsets of volumes from the
macaque dMRI corresponding to different shells. Details for each combination
are shown in Table 1. We fitted the ball-and-stick model6,7 to
each dataset and performed probabilistic tractography in FSL, using the
injection site as seed. Chemical tracing revealed three bundles of frontal pole
fibers (Fig.1): uncinate fasciculus-UF, corpus callosum-CC and internal
capsule-IC. A connection present in both dMRI tractography and tracer data
was deemed true positive (TP); one present only in tractography (e.g., the
fornix) was deemed a false positive (FP).
Human
data: We used in-vivo human dMRI data from the MGH/UCLA
Human Connectome Project (3T Siemens Skyra, maximum gradient=300mT/m, 512
directions, bmax=10000, 1.5mm resolution). We extracted different subsets of
volumes, listed in Table 1. Seeds for tractography were placed in the region
homologous to the macaque injection site. Taking advantage of inter-species homologies
in frontal cortex4, we
deemed streamlines going through the UF, CC, IC as TPs and all others (e.g.,
the fornix) as FPs.
Results
We
evaluated tractography accuracy on each dataset by computing the TP and FP rate
as the fraction of connections reconstructed by tractography that are,
respectively, present and absent in the ground truth. We repeat this for
varying levels of thresholding on the probabilistic tractography map to plot receiver-operating
characteristic (ROC) curves. Fig.3 and 4 show ROC curves for each acquisition
scheme of the monkey and human datasets respectively.
Discussion
Our
findings show that there are diminishing returns, in terms of tractography
accuracy, above a certain maximum b-value. As seen in Fig. 3, accuracy in the
ex-vivo macaque brain does not improve substantially by adding shells beyond a
maximum b-value of 25.6k and 256 directions, and, in some case, becomes worse.
Note that this would be equivalent to a b-value of 6K-7K in vivo, as
diffusivity in the fixed macaque brain is at 25% of its in vivo values8. As
seen in Fig. 4, accuracy in the in vivo human data for reconstructing the
homologous projections did not improve substantially when adding the b=10K
shell.
Conclusion
We used objective, quantitative metrics to
compare dMRI acquisition schemes in terms of the accuracy of tractography
originating in BA10. Our results indicate that there is an upper bound to the
angular resolution of dMRI, beyond which accuracy does not improve.
Encouragingly, the optimum is within the range feasible in vivo with current
state-of-the-art ultra-high diffusion gradients and accelerated sequences.
These results are specific to the accuracy of frontal pole projection, and the
analyses methods used here. Our future work will analyze pathways from
additional injection sites to determine whether optimal acquisition parameters
are global or depend on the brain network of interest. We will also investigate
optimal acquisition schemes for other diffusion model-fitting and tractography
methods.
Acknowledgements
This research was carried out in whole or in part at the Athinoula A. Martinos Center for Biomedical Imaging at the Massachusetts General Hospital, using resources provided by the Center for Functional Neuroimaging Technologies, P41EB015896, a P41 Biotechnology Resource Grant supported by the National Institute of Biomedical Imaging and Bioengineering (NIBIB), National Institutes of Health and the NIH Blueprint for Neuroscience Research (T90DA022759/R90DA023427) This work also involved the use of instrumentation supported by the NIH Shared Instrumentation Grant Program and/or High-End Instrumentation Grant Program; specifically, grant number(s) S10RR016811, S10RR023401, S10RR019307, S10RR019254, S10RR023043. References
1 Thomas, C. et al.
Anatomical accuracy of brain connections derived from diffusion MRI
tractography is inherently limited. Proc.
Natl. Acad. Sci. 111,
16574–16579 (2014).
2 Gao, Y. et al.
Validation of DTI tractography-based measures of primary motor area
connectivity in the squirrel monkey brain. PLoS
One 8, e75065 (2013).
3 Dauguet, J. et al.
Comparison of fiber tracts derived from in-vivo DTI tractography with 3D
histological neural tract tracer reconstruction on a macaque brain. Neuroimage 37, 530–538 (2007).
4 Jbabdi, S., Lehman,
J. F., Haber, S. N. & Behrens, T. E. Human and monkey ventral prefrontal
fibers use the same organizational principles to reach their targets: tracing
versus tractography. J. Neurosci.
33, 3190–201 (2013).
5 Schmahmann, J. D. et al. Association fibre pathways of the
brain: Parallel observations from diffusion spectrum imaging and
autoradiography. Brain 130, 630–653 (2007).
6 Van Essen, D. C. et al. Diffusion
MRI. Diffusion MRI
(Elsevier, 2014). doi:10.1016/B978-0-12-396460-1.00016-0
7 Jbabdi, S.,
Sotiropoulos, S. N., Savio, A. M., Graña, M. & Behrens, T. E. J.
Model-based analysis of multishell diffusion MR data for tractography: How to
get over fitting problems. Magn. Reson. Med.
68, 1846–1855 (2012).
8 D’Arceuil, H. E.,
Westmoreland, S. & de Crespigny, A. J. An approach to high resolution diffusion
tensor imaging in fixed primate brain. Neuroimage
35, 553–565 (2007).