On estimation of bias field in MRI images

Abstract

Bias field or Intensity inhomogeneity (IIH) or intensity non-uniformity (INU) in magnetic resonance imaging (MRI) is an artifact that is mainly produced by improper image acquisition process. Bias field, which is slow variant in nature, affects the intensities of the homogeneous tissue regions (for example, gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) in MRI brain images). In this paper, we present a novel retrospective bias-correction technique using polynomial surface fitting. The input image is segmented into different homogeneous regions by considering intensity histogram of the corresponding MRI slice. The proposed method works on each segment individually to estimate the approximated bias-field using second-order and third-order polynomial surface fitting method. The bias fields obtained from the homogeneous tissue regions are ensemble to estimate the approximate inhomogeneity map of the entire image. We obtained the bias-corrected image after removing the bias-field iteratively. Comparative study and quantitative evaluation of the proposed second-order and third-order surface fitting methods on MRI brain images were performed by comparing standard deviations of different homogeneous tissue regions. The simulation results show that in most of the cases the second-order polynomial outperforms the third-order polynomial for estimating the bias fields in MRI images.