Premature births, or births before 37 gestational weeks, are associated with structural and functional abnormalities in the brain. Magnetic Resonance Imaging (MRI) is used to examine brain injury in neonates. Previous work demonstrated correlations between neonatal cortical lobes volumes from MRI and scores on the Bayley Scales of Infant and Toddler Development-Third Edition (Bayley-III), which measures cognitive, motor, and language skills of infants. The occipital lobe is important to diagnosis of visual and neurodevelopmental delays. In this work, I examined the relationship between occipital lobe subdivision volumes and Bayley-III scores. In order to do this, I designed a parcellation or labeling protocol that divided the occipital lobe into two regions based on the calcarine sulcus, the cuneus region and the lingual region. I applied the protocol to the 32 MRI training dataset. Using this parcellation protocol, I created a spatio-temporal atlas of brain parcels, which is a time and space model of brain tissue and parcel probability estimates. This atlas was used to validate our protocol using a leave-one-out validation and to automatically parcellate the larger dataset using EM or Expectation-Maximization labeling. The dataset contained 270 T1-weighted MRI scans from premature neonates who took the Bayley-III at 18 months. Once the larger dataset was labeled, I used multi-variate linear regression, normalizing for head size, to examine the relationship between occipital lobe gray matter volumes and scores on the Bayley-III and analyzed the growth pattern of the regions of the occipital lobe. Both regions showed symmetric and exponential growth patterns. Statistically significant correlations between the lingual region of the occipital lobe and Bayley-III scores were found. This suggests that regional occipital lobe volumes are associated with developmental outcomes in premature neonates.

The study of neonatal neuroanatomy is a complex field because of the rapid changes occurring continuously. The physical development of tissues is critical to understanding the cognitive development of behavior and learning in the neonate. These structures can be visualized and analyzed with Magnetic Resonance Imaging (MRI). The hippocampal structure in particular is vital to memory consolidation and is therefore key to understanding how newborns build upon acquired knowledge and how this affects their behavior as they grow. This strucure has been studied at many different ages with parallel work analyzing how changes in anatomy result in observable cognitive differences and developments. This study in particular has a rare cohort of scans that were taken shortly after birth. Prematurity is an especially important factor to consider here as it affects physical and behavioral growth. Here we looked at 42 premature MRI subjects and used these to build a temporal model of brain development. Our primary aims of the study are to: (1) develop a protocol for labeling the hippocampus in premature neonate scans, (2) create a growth over time model of the structure in order to automatically segment a larger number of scans and (3) observe how volume varies with behavior as projected by Bayley's Scale of Infant and Toddler Development. So far aim 1 has been achieved and current work is increasing the previous atlas size of 32 scans to 42 in order to improve automatic segmentation. The overarching goal of premature neonatal MRI segmentation is to provide further data for early diagnosis of behavior abnormalities that could develop later in the subject. Results here could produce a meaningful indicator of cognitive development through MRI analysis as well as assist in improving the accuracy of MRI automatic segmentation algorithms.