Hybrid Method for Brain Tissue Classification in Magnetic Resonance Images of Children

Abstract

Background: Automatic brain tissue classification in children is tedious owing to motion artifacts, rapid brain maturation, etc. Considering this, we propose a hybrid approach for tissue classification. Methods: This approach utilizes atlas-free technique for brain extraction and labeling of Cerebrospinal Fluid and subsequent atlas-based method for classification of Gray and White Matter tissues. The results for brain, Gray Matter, White Matter and Cerebrospinal Fluid are validated quantitatively using Dice ratio. Findings: Mean Dice values of 0.9767 for brain, 0.8503 for Gray Matter, 0.7850 for White Matter and 0.8224 for Cerebrospinal Fluid are achieved. This signifies excellent agreement since Dice values for all tissues are much above the accepted similarity level of 0.7. Applications: The automatic tissue classification can be used to carry out volumetric tissue computations to understand brain maturation in childhood.