Novel vector-valued approach to automatic brain tissue classification

Abstract

In this work we propose a novel SSIM (Structural Similarity Index Measure)-guided brain tissue classification approach, implementing Kernel Fisher Discriminant Analysis (KFDA). In Computer Vision, KFDA has been shown to be competitive with other state-of-the-art techniques. In the KFDA-based framework, we exploit the complex structure of grey matter, white matter and cerebro-spinal fluid intensity clusters to find an optimal classification. We illustrate our novel technique using a dataset of early normal brain development in the age range from 10 days to 4.5 years. The SSIM metric, an objective measure of an image quality as perceived by the Human Visual System, is used to evaluate the quality of brain segmentation. SSIM comparison of the quality of classification obtained by the KFDA-based and the Expectation-Maximization algorithms shows the superior performance of the proposed technique. © 2013 Springer-Verlag.