1030 - Probabilistic Neighborhood Tractography in the Preterm Neonatal Brain

Introduction/target audience: Preterm birth is a leading cause of cognitive impairment in childhood and is associated with alterations in brain
development that are apparent in the neonatal period. Brain structural changes associated with preterm birth include enlargement of the ventricular
system, reduced cortical complexity and diffuse white matter signal abnormalities on structural MRI.1
Diffusion MRI (dMRI) and tractography may
provide further insights into the cerebral microstructural changes that accompany preterm birth by supplying quantitative biomarkers of white matter
integrity in specific tracts of interest.2
Purpose: This pilot work describes the first application of an automatic single seed point tractography-based segmentation method, probabilistic
neighborhood tractography (PNT),3,4 to the study of the preterm brain. PNT, which can segment the same fasciculus across groups of subjects and
provide quantitative measures of tract integrity and shape, works by placing seed points in a neighborhood surrounding a seed point transferred from
standard space, with the tract that best matches a predefined reference tract in terms of length and shape chosen from this group of ‘candidate’ tracts.3
Methods: Twelve preterm infants with mean postmenstrual age (PMA) 28.43 weeks (range 25.71-30.14) underwent a high angular resolution axial
dMRI protocol at term equivalent age (mean PMA 40.28 weeks [range 38-42.56]) without sedation, and with informed parental consent and appropriate
ethical approval. The dMRI protocol, acquired using a MAGNETOM Verio 3 T clinical scanner (Siemens AG, Healthcare Sector, Erlangen, Germany),
consisted of 11 T2- and 64 diffusion-weighted (b = 750 s/mm2
) single-shot, spin-echo, echo planar imaging volumes acquired with 2 mm isotropic
voxels (field of view 256 × 256 mm, imaging matrix 128 × 128, 50 contiguous interleaved slices with 2 mm thickness).
Eight tracts-of-interest were identified using PNT from the dMRI data as implemented in the TractoR package for fiber tracking analysis
(http://www.tractor-mri.org.uk).4
Tracts assessed were the genu and splenium of corpus callosum, cingulum cingulate gyri (CCG), left and right
projections of the corticospinal tract (CST), and inferior longitudinal (ILF) fasciculi. Using a 7 × 7 × 7 neighborhood of seed voxels, the seed point
which produced the best matching tract to the reference in terms of both length and shape was determined using tract shape models determined from a
separate group of normal volunteers aged 25 to 65 years. To reduce false positives and reduce noise-related fall-off in connection probability with
distance from the seed point, pruning of streamlines that did not resemble the median path of the best match tract was employed.
4
In a further step, all
best match tracts were visually assessed by an experienced rater, and any subject with aberrant or truncated pathways that were not anatomically
plausible representations of the fasciculi-of-interest excluded from further analysis. For anatomically acceptable tracts, the resulting tractography masks
were applied to each participant’s mean diffusivity (〈D〉), fractional anisotropy (FA), axial (λAxial) and radial (λRad) diffusivity volumes, which
permitted tract-specific mean values of these biomarkers, weighted by the connection probability, to be determined for each tract in every subject.
Finally, the absolute goodness-of-fit of the best match tract to the reference (R) for each subject was determined from the log-ratio between the
matching likelihood of the chosen candidate tract and the matching likelihood of the reference tract to itself.3,5 Since the reference tract has, by
definition, a log-ratio of zero, this measure of topological similarity will almost always be negative for all other tracts; and the more negative it is, the
less good is the fit between the reference and best match tract.
Results: Figure 1 shows the tract segmentation across all 12 subjects for genu and splenium, and
indicates the close spatial correspondence of the segmented pathways for these two tracts. Visual
assessment of the individual segmented tracts indicated that PNT provided anatomically
acceptable representations of the fasciculi of interest for the vast majority of pathways (92 % over
all subjects and tracts), with a minimum of 75 % for right CCG.
Mean (± SD) values of tract-averaged 〈D〉, FA, λAxial and λRad for the eight fasciculi of interest
are presented in Table 1. Values of 〈D〉 range from 1139 ± 70 for right CST to 1707 ± 209 μm2
/s
for left ILF, while FA ranges from 0.19 ± 0.02 in left ILF to 0.31 ± 0.03 in splenium. Values of
λAxial vary from 1512 ± 76 for right CST to 2061 ± 161 μm2
/s for splenium, while λRad ranges
from 952 ± 87 for right CST to 1532 ± 200 μm2
/s for left ILF. Finally median (± IQR/2) values of
R range from -3.68 ± 0.79 for genu to -47.26 ± 7.45 for left CST and are generally lower, i.e.
showing less topological similarity to the reference tract, than those seen in the adult brain.5
Discussion/Conclusions: These pilot data show for the first time that quantitative
measurements of dMRI biomarkers can be made in the preterm brain from high angular
resolution dMRI data using PNT. These values are comparable to other studies using
tractography methods and demonstrate the increased diffusivities and reduced FA indicative of white matter development at this point in early life
compared with the adult brain.2
Of particular interest is the fact that the method is able to identify successfully a range of fasciculi using reference tracts
obtained from the adult brain. We are currently investigating whether the use of reference tracts from infants further
improves this method, and whether the tract shape parameter R provides additional useful information about brain
structure that can be used to assess cerebral development in preterm birth and potential therapeutic interventions.