671 - Diffusion Tensor Imaging Brain Atlas-based White Matter Analysis in Preterm Infants

Publication Number: 2654.671

Hideyoshi Fukui, Department of neonatology, Izumiotsu Women's and Children's Hospital, Osaka Metropolitan University Graduate School of Medicine, Izumiotsu-shi, Osaka, Japan; Jingang Li, Izumiotsu Women's and Children's Hospital, Izumiotsu, Osaka, Japan; Kentaro Yasuhira, Department of neonatology, Izumiotu Women’s and Children's Hospital, Izumiotsu-shi, Osaka, Japan; Nagisa Yamamoto, Department of neonatology, Izumiotsu Women's and Children's Hospital, izumiotsu-shi, Osaka, Japan; Shuichi Takimoto, Department of radiology,Izumiotsu Women's and Children's Hospital, Izumiotsu-shi, Osaka, Japan; Daisaku Kitano, Department of radiology, Izumiotsu Women's and Children's Hospital, Izumiotsu-shi, Osaka, Japan; Masao Hamuro, Izumiotsu Women's and Children's Hospital, Sakai, Osaka, Japan; Ayumi Murayama, Yodogawa Christian Hospital, Osaka, Osaka, Japan; Nanae Yutaka, Yodogawa Christian Hospital, Osaka, Osaka, Japan; Yumiko Uchida, Nara Medical University, Kashihara, Nara, Japan; Toshiya Nishikubo, Nara Medical University Hospital, Nara, Nara, Japan; Touhei Usuda, Niigata City General Hospital, Japan, Niigata, Niigata, Japan; Takashi Sato, Niiagta City general hospital, Niigata, Niigata, Japan; Chao-Hsiung Hsu, Howard University College of Medicine, College Park, MD, United States

Background: Preterm infants are frequently complicated with white matter injury (WMI). Although conventional T1/T2-weighted MRI has limitation in evaluating WM, diffusion tensor imaging (DTI), which reveals the diffusion of water molecules, allows the assessment of WM structure. Recently, MRI atlas-based analysis (ABA) has widespread in neuroscience and neurology research. In particular, DTI-based ABA allows sensitive detection of WM structure, facilitates more accurate parcellation and evaluation of WM regions than conventional MRI-based ABA.

Objective: Conducting comparative analysis of WM regions between term and preterm infants by DTI-ABA for quantitative detection of WMI.

Design/Methods: Isovoxel (resolution: 2 mm) and gapless DT images (b values = 0 and 1000, 12 directions) were acquired using 1.5T-MRI scanner from 81 infants at term-equivalent age. Template DTI-based neonatal brain atlas (Oishi, 2011) was transformed to fit each subject DT image using DiffeoMap (John's Hopkins University), parcellating into 82 WM regions. The values of fractional anisotropy (FA) and radial diffusivity (RD) of the 82 regions were measured, and then categorized into nine WM regions. We also categorized infants into three groups: born at term, 28-33 weeks gestational age (GA) (very preterm infants) and 23-27 weeks GA (extremely preterm infants), and statistically compared between the groups.

Results: There was no significant difference in brain volume among the three groups. Very preterm infants showed significantly lower FA in corpus callosum (CC), internal capsule (IC), corona radiata (CR) and all cortical WM, and higher RD in CC, IC, CR, and frontal, parietal and occipital WM compared with term infants. Extremely preterm infants showed lower FA in all cortical WM except for parietal region, and higher RD in parietal and occipital WM compared with very preterm infants (P < 0.05).

Conclusion(s): DTI-ABA revealed significantly lower FA and higher RD across cortical WM regions in more premature infants, indicating more immature brains develop more severe dysmaturation or impairment of myelination, specifically in the peripheral WM. Findings from the present study were partly consistent with previous reports analyzed with conventional methods, validating the accuracy of DTI-ABA. This study demonstrated DTI-ABA allows reliable automatic parcellation of the whole brain and inter-group regional quantitative comparisons, facilitating effective detection of regional alterations particularly in the WM, providing new insight in the preterm brain.

Workflow of image and atlas registration for automated anatomical labeling
DiffeoMap (Johns Hopkins University) was used to first linear normalize each subject FA and b0 brain images to the template FA and b0 common space, respectively, followed by a non-rigid registration (Large deformation diffeomorphic metric mapping, LDDMM) to obtain the registration matrices. Then, neonatal brain atlas (Oishi, 2011) was transformed to fit each subject brain by inverse registration (LDDMM followed by inverse linear registration) using the matrices obtained, and superimposed on the original subject image, allowing parcellation of 122 regions. Fifty cortical regions were then separated into 50 cortical GM and WM regions by threshold of FA 0.2, ending up with automatically parcellating in total 82 WM regions from the whole brain.

FA values at 82 WM regions
Very preterm infants showed significantly lower FA values in corpus callosum (CC), internal capsule (IC), corona radiata (CR) and all cortical WM compared with term infants. Extremely preterm infants showed significantly lower FA values in all cortical WM except for parietal region, compared with very preterm infants.

RD values at 82 WM regions
Very preterm infants showed significantly higher RD in CC, IC, CR and FG-, PL- and OG-WM compared with term infants. Extremely preterm infants showed significantly higher RD in PL- and OG-WM compared with very preterm infants.