Skull and scalp segmentation in neonatal cerebral MRI using subject-specific probability models

Abstract

This study presents a new approach for segmenting cranial bones in magnetic resonance images (MRIs) acquired from neonates in the gestational age range of 39 to 42 weeks. the approach uses subject-specific probability maps of the skull and scalp, created from atlas computed tomography (CT) images taken retrospectively from neonates in the same age range. the method also uses a subject-specific probability map of cerebrospinal fluid (CSF), constructed from retrospective atlas MRIs. To build skull, scalp, and CSF probability maps, a subject-specific bimodal MR-CT neonatal head template is employed. In the next step, the subject-specific probability maps are fed to the expectation maximization algorithm in conjunction with Markov random field method implemented in FSL software to segment the skull and scalp from the input MR image. The results of the proposed method were evaluated through various experiments. First, we employed our method as a brain tissue extractor and compared its results with public methods such as the Brain Extraction Tool (BET) and Brain Surface Extractor (BSE). Second, we calculated the similarity in shape between the frontal and occipital sutures (which had been reconstructed from segmented cranial bones) and the ground truth. For this purpose, modified versions of the Dice similarity coefficient (DSC) were adopted and used. Finally, retrospective data including MRI and CT images obtained from the same neonate within a short time interval were used. After aligning the two images, the DSC and modified Hausdorff distance (MHD) were used to compare the similarity of the cranial bones in the MR and CT images. Furthermore, the anterior fontanel size was compared to the normal size reported for neonates in the same age range. Cranial bone thickness was calculated and compared to normal values reported for healthy neonates. The results of these experiments demonstrated the success of our segmentation method. The algorithm for creating subject-specific atlases is publicly accessible through a graphical user interface at medvispy.ee.kntu.ac.ir.