Retrospective correction of MR intensity inhomogeneity by information minimization

Abstract

In this paper, the problem of retrospective correction of intensity inhomogeneity in magnetic resonance (MR) images is addressed. A novel model-based correction method is proposed, based on the assumption that an image corrupted by intensity inhomogeneity contains more information than the corresponding uncorrupted image. The image degradation process is described by a linear model, consisting of a multiplicative and an additive component which are modeled by a combination of smoothly varying basis functions. The degraded image is corrected by the inverse of the image degradation model. The parameters of this model are optimized such that the information of the corrected image is minimized while the global intensity statistic is preserved. The method was quantitatively evaluated and compared to other methods on a number of simulated and real MR images and proved to be effective, reliable, and computationally attractive. The method can be widely applied to different types of MR images because it solely uses the information that is naturally present in an image, without making assumptions on its spatial and intensity distribution. Besides, the method requires no preprocessing, parameter setting, nor user interaction. Consequently, the proposed method may be a valuable tool in MR image analysis.

CAIMAN: An online algorithm repository for Cancer Image Analysis

“…Up to the authors' best knowledge, only one webpage created by Likar and co-authors [18] provides a similar approach to CAIMAN. This webpage processes images for intensity inhomogenetiy correction using an information minimization algorithm [19]. The webpage requires the user to provide an image in raw format and the following information X,Y,Z dimensions, X,Y,Z voxel dimension, bits per voxel, byte order (little endian/big endian), thresholding and subsampling parameters as well as the correction model.…”

The fast automatic algorithm for correction of MR bias field

“…However, as noted in Ref. 15, it does not provide the information about the overlap between intensity distributions of distinct tissue classes. A bias correction method may transform an image in such a way that coefficients of variation for white and GM will be reduced but the intensities of WM and GM will become more difficult to distinguish.…”

“…It is common for such methods to construct a target function and iteratively minimize it to find the basis coefficients (15)(16)(17). Fuzzy clustering methods (3,18,19) are hybrids of surface fitting and EM algorithms.…”

Computer aided diagnosis in chest radiography

“…Many image acquisition systems contain artifacts that are hard to correct with calibration schemes. One example in medical image processing is the inhomogeneity of the magnetic field of an MR scanner or of the sensitivity of MR surface coils, leading to low frequency gradients over the image, for which retrospective correction schemes have been proposed [156,157]. A generated example is shown in Figure 4.7 where we multiplied a texture image with Gaussian noise.…”