Retrospective Evaluation and Correction of Intensity Inhomogeneity - Advance Image Processing in Magnetic Resonance Imaging - page 140

5

Retrospective Evaluation
and Correction of
Intensity Inhomogeneity

Martin Styner and Koen Van Leemput

CONTENTS

5.1 Introduction..............................................................................................145
5.2 Early Solutions ........................................................................................148
5.3 Combined Segmentation and Inhomogeneity
Correction Methods .................................................................................149
5.4 Correction Based on Evaluation of the Histogram.................................160
5.4.1 Parametric Bias Correction .........................................................160
5.4.2 Information Minimization and N3 ..............................................163
5.5 Discussion and Conclusion .....................................................................164
Acknowledgments .............................................................................................165
References .........................................................................................................165

5.1 INTRODUCTION

Image intensity inhomogeneity, also called intensity nonuniformity (INU), in
magnetic resonance imaging (MRI) is an adverse effect that affects qualitative
and quantitative analysis of the images. The inhomogeneities are characterized
by slowly varying intensity values of the same tissue over the image domain. The
variability of the tissue intensity values with respect to image location can severely
affect visual evaluation as well as the result of image processing algorithms based
on absolute intensity values. The effect of the intensity inhomogeneities on the
subsequent segmentation has been observed and discussed already in early appli-

cations (e.g., [1-5]; see also Figure 5.1 ). It has also become a highly important

issue in quantitative morphometry, which is either based on volumetric measure-
ments from segmentation or residual image differences after registration. There are
numerous sources for the emergence of MRI intensity inhomogeneity. Some of
them can be overcome by regular calibration, whereas many of them are different
from scan to scan, for example, when inhomogeneities are due to geometry and

145