Fig 1.
(a) Construction of an anatomical atlas from segmented fibers. (b) Extraction and estimation of the TOD anatomical prior from the atlas. (c) Combination between the prior and the subject data using a Riemannian framework.
Fig 2.
Examples of combination of the data and the prior.
On the first row: prior and data are aligned. The EODF main orientation are unchanged. On the second and third rows: Prior and data are crossing. The corresponding EODF peak is more or less attenuated using the weight of the prior.
Fig 3.
Dice score for various α values (from the Eq 8) for the tracking of the CST and for the iFOD and PTT algorithms.
The method performs most effectively with an α value of 0.35.
Fig 4.
Right: Prior weighting map example for the DiSCo dataset. Higher value denotes higher usage of the prior and thus more complex regions.
Fig 5.
Pearson correlation for the DiSCo experiments:
(a) Fiber pathways ground-truth experiment with respect to SNR (b) Multi-Shell ODF experiment for the three DiSCo datasets. First column for PTT and second for iFOD2. In all cases, blue points denotes measure without prior and the oranges ones with prior.
Fig 6.
Visual comparison of the resulting tractograms. From top to bottom line: iFOD2, PTT, iFOD2 with prior, PTT with prior, iFOD2 with BST priors and PTT with BST priors.
Last line: Prior weighting map, where brighter pixel denotes a higher usage of the prior. From left to right: The Arcuate fascicle (AF), the Cingulum (CG), the Corticospinal tract (CST), the Optic radiation (OR) and the Superior longitudinal fascicle I (SLF I). Note that the quantitative results presented in the rest of the work use both sides of each bundles. But to simplify visualization, only one side is shown here for all bundles, except for the CST.
Fig 7.
On the left, Dice score result for the HCP experiment.
On the right, Valid Streamline (VS) score result for the HCP experiment. From top to bottom : The Arcuate fascicle (AF), the Cingulum (CG), the Corticospinal tract (CST), the Optic radiation (OR) and the Superior longitudinal fascicle I (SLF I).
Fig 8.
Comparison with the BST algorithm for the AF and CST bundles.
The top row present the Dice score and the bottom row the VS numbers. In average, our method preforms better for both measures.