An automatic pipeline for atlas-based fetal and neonatal brain segmentation and analysis

TL;DR

Findings show the potential of the presented atlases and the whole pipeline for application in both fetal, neonatal, and longitudinal studies, which could lead to dramatic improvements in the understanding of perinatal brain development.

Abstract

The automatic segmentation of perinatal brain structures in magnetic resonance imaging (MRI) is of utmost importance for the study of brain growth and related complications. While different methods exist for adult and pediatric MRI data, there is a lack for automatic tools for the analysis of perinatal imaging. In this work, a new pipeline for fetal and neonatal segmentation has been developed. We also report the creation of two new fetal atlases, and their use within the pipeline for atlas-based segmentation, based on novel registration methods. The pipeline is also able to extract cortical and pial surfaces and compute features, such as curvature, thickness, sulcal depth, and local gyrification index. Results show that the introduction of the new templates together with our segmentation strategy leads to accurate results when compared to expert annotations, as well as better performances when compared to a reference pipeline (developing Human Connectome Project (dHCP)), for both early and late-onset fetal brains.

Figures (12)

• Figure 1: Perinatal pipeline workflow.
• Figure 2: Neonatal and Fetal temporal templates with corresponding original segmentations. (a) Neonatal template at 28, 36 and 44 weeks of gestation respectively. (b) Fetal template at 19, 29 and 39 weeks of gestation respectively.
• Figure 3: The process to generate the multi-subject fetal atlas starts from the reconstructed T2-weighted images. A pre-processing phase (a) consisting of N4 bias correction and brain extraction is performed first, followed by the tissue segmentation (b) using the modified fetal template and a manual refinement. From the obtained segmentation, gray matter and ventricles segmentation are extracted (c). They are used in a three-channel registration with the ALBERTs atlas (d) to obtain a first estimate of their structural segmentation and generate the multi-subject atlas (e). Each subject is then segmented iteratively using the other 19 subjects (f) of the atlas until convergence, and at each step the atlas is updated (g).
• Figure 4: (a) Registration phase as performed in the dHCP pipeline: the subject (left) is registered to each one of the subject composing the multi-subject atlas; (b) Registration phase as performed in the proposed pipeline: the subject is registered to the computed template, and the resulting transformation combined with the previously stored transformations between the template and the subjects composing the multi-subject atlas.
• Figure 5: Temporal fetal template with the equivalent 7-tissue segmentation.