Parcellation of neonatal brain MRI into 107 regions using atlas propagation through intermediate time points in childhood.

Manuel Blesa¹, Ahmed Serag¹, Alaistir G Wilkinson², Devasuda Anblagan^1,3, Emma J Telford¹, Rozalia Pataki¹, Sarah A Sparrow¹, Gillian Macnaught⁴, Scott I Semple⁴, Mark E Bastin³, and James P Boardman^1,3

¹MRC Centre for Reproductive Health, University of Edinburgh, Edinburgh, United Kingdom, ²Department of Radiology, Royal Hospital for Sick Children, Edinburgh, United Kingdom, ³Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom, ⁴Clinical Research Imaging Centre, University of Edinburgh, Edinburgh, United Kingdom

Synopsis

We created a neonatal brain atlas of healthy subjects that can be applied to multi-modal MRI data. Structural and diffusion 3T MRI scans were acquired after birth from 25 neonates born at term. The SRI24/TZO atlas was propagated to the neonatal data using temporal registration via childhood templates (NIHPD), with the final atlas (the Edinburgh Neonatal Atlas, ENA25) constructed using iterative averaging of T1-weighted volumes. The computed transformations were applied to T2-weighted data, diffusion maps and tissue probability maps to provide a multi-modal atlas with 107 anatomical regions; and we have generated a symmetric version to facilitate studies of laterality.

Introduction

Neuroimage analysis pipelines rely on parcellated atlases generated from healthy individuals to provide anatomic context to structural and diffusion MRI data (sMRI, dMRI). Atlases constructed using adult data introduce bias into studies of early brain development¹. The aim of this work is to create a neonatal brain atlas of healthy subjects that can be applied to multi-modal MRI data, and could serve as a useful start point for modelling brain growth during development.

Materials and methods

Participants: Twenty-five healthy control infants (12 males and 13 females, mean post-menstrual age (PMA) at birth 39⁺⁴ weeks, range 37⁺²-41⁺²) underwent brain MRI at mean age 42.22 weeks (range 39-47⁺¹). MRI acquisition: Imaging was conducted at the Clinical Research Imaging Centre, Edinburgh, UK, on a Siemens MAGNETOM Verio 3T MRI clinical system: 3D T1-weighted (T1w) MPRAGE with voxel size = 1 x 1 x 1 mm; T2-weighted (T2w) SPACE with voxel size = 0.9 x 0.9 x 0.9 mm; dMRI using a protocol consisting of 11 T2- and 64 diffusion-weighted (b = 750 s/mm2) single-shot spin-echo echo-planar imaging volumes acquired with matrix = 128 × 128 and 50 contiguous interleaved slices with 2 mm thickness. Atlas construction: The atlas construction framework consisted of two main steps. First, the neonatal brain template of the NIH Pediatric Database (NIHPD)² was parcellated into anatomical regions of interest (ROIs) using temporal registration^3,4 of an adult atlas (SRI24)⁵ via 7 intermediate spatio-temporal templates from the NIHPD. Second, the parcellated NIHPD neonatal brain template was propagated to a cohort of term-born neonates, and a groupwise atlas constructed using an iterative averaging approach⁶. All transformations were computed using the T1w volumes and the resulting transformations were applied to T2w, FA and MD volumes, the corresponding label maps, and tissue segmentation (WM, GM, CSF)^7,8. Finally, to create a symmetric version of the template, we flipped each subjects’ T1w volume left to right, and using each volume as an independent subject in the template creation². Visual Inspection and correction: Labels were inspected and edited where necessary by a radiologist experienced in neonatal brain MRI (A.G.W.).

Results

The constructed atlas (Fig. 1) consists of 107 anatomical regions and provides different (Fig. 2) that work across different MRI modalities. Because of known asymmetries in the brain⁹, we created a symmetric version of the atlas which can be used to investigate laterality in the developing brain (Fig. 2).

Discussion

We created a neonatal atlas (ENA25) that parcellates the brain into 107 anatomical regions of interest. The atlas contains templates (symmetric and asymmetric) for different MR modalities [T1w, T2w, diffusion tensor MRI (FA and MD)] and tissue probability maps. The atlas can be used to perform different studies that would benefit from an age specific template with a large number of labels like volumetric studies of different tissues or regions of interest, structural connectivity of the neonatal brain or tract-based studies¹⁰. To our knowledge, the constructed atlas has the greatest anatomical detail of the neonatal brain, and because the atlas is generated from step-wise propagation of adult labels through intermediate time points in childhood, it may serve as a useful start point for modelling brain growth during development.

Acknowledgements

We are grateful to the families who consented to take part in the study and to the nursing and radiography staff at the Clinical Research Imaging Centre, University of Edinburgh (http://www.cric.ed.ac.uk) who participated in scanning the infants. The study was supported by Theirworld, NHS Research Scotland, and NHS Lothian Research and Development. We thank Thorsten Feiweier at Siemens Healthcare for collaborating with dMRI acquisitions (Works-in-Progress Package for Advanced EPI Diffusion Imaging).

References

[1] Wilke, M. et al. Normative pediatric brain data for spatial normalization and segmentation differs from standard adult data. Magn Reson Med, 2003. [2] VS Fonov, et al. Unbiased nonlinear average age-appropriate brain templates from birth to adulthood. NeuroImage, 2009 [3] Serag, A. et al. LISA: Longitudinal image registration via spatio-temporal atlases. Biomedical Imaging (ISBI), 2012 [4] Avants, BB. et al. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain Med Image Anal, 2008 [5] Rohlfing, T. et al. The SRI24 multichannel atlas of normal adult human brain structure. Hum Brain Mapp, 2010. [6] Guimond, A. et al. Average Brain Models: A Convergence Study. Computer Vision and Image Understanding, 2000. [7] Serag, A. et al. Construction of a consistent high-definition spatio-temporal atlas of the developing brain using adaptive kernel regression. NeuroImage, 2012 [8] Serag, A. et al. A Multi-channel 4D Probabilistic Atlas of the Developing Brain: Application to Fetuses and Neonates. Annals of the BMVA, 2012 [9] Dubois, J. et al. Structural asymmetries of perisylvian regions in the preterm newborn. NeuroImage, 2010 [10] Anblagan, D. et al. Tract shape modeling detects changes associated with preterm birth and neuroprotective treatment effects. NeuroImage: Clinical, 2015.

Figures

Figure 1: 3D rendered volume.

Figure 2: From left to right: T2w, T1w and label parcellation map (the asymmetric template), T1w and label parcellation map (the symmetric template), FA and probability maps for tissue segmentation (GM and WM). MD template and CSF probability map are not shown.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)

4421