Polynomial surface fitting based method for retrospective correction of intensity inhomogeneity in MR images

Abstract

In this paper we present a polynomial surface fitting based retrospective correction method for intensity inhomogeneity in Magnetic Resonance (MR) images. The intensity inhomogeneity, which is also known as bias field or intensity non-uniformity, may arise due to erroneous acquire process causes severe image degradation. The retrospective methods are capable of correcting intensity nonuniformity and are widely applied to both simulated and real MR volumes. The polynomial surface fitting based retrospective intensity correction technique works on the assumption that the large homogeneous tissues (e.g. Gray matter, white matter, etc.) are distributed throughout the image so that a parametric polynomial based surface can be fitted on image features to estimate the intensity inhomogeneity map. The image features which contain information about intensity inhomogeneity are generally image intensity values or intensity gradient values. The bias field, which is characterized as a slow varying function, is further used to obtain the corrected image. The quantitative evaluation of the correction strategy is done by comparing the standard deviation between input and output images which is expected to be reduced in principle tissues. Simulation results on synthetic MR images show that the proposed method can eliminate intensity inhomogeneity quite satisfactorily.