Effectiveness of Intensity Normalization on Human Brain MRIs with Multiple Sclerosis

Tissue intensity standardization is an important preprocessing step in the study and analysis of Magnetic Resonance Images (MRI) of human brain. Sources of variations in the intensity ranges across different MRI volumes, even after intensity inhomogeneity correction, can result from heterogeneity of data due to difference in scanners, presence of Multiple Sclerosis (MS) lesions and the stage of disease progression in the brain. As a result, intensity normalization plays a significant role in standardizing the tissue intensity ranges across MRI volumes on which most automatic image analysis methods base their distributional assumptions. The method of Nyul et. al. [3] has become a widely used standard for intensity normalization. However, an extensive validation of this approach on multi-site multi-scanner data in the presence of MS has yet not been performed. We aim to undertake this validation in this work and show the effectiveness of this procedure on standardizing tissue intensities in the MRI volumes.