A Model-Independent, Multi-Image Approach to MR Inhomogeneity Correction

We present a technique for model-independent estimation of the inhomogeneity effect across multiple different acquisitions of the same subject, on the assumption of equal inhomogeneities in each image. The technique operates by integration of smoothed estimates of the local derivative of the bias field. Combination of evidence from multiple images is achieved at the derivative estimation stage via statistical weighting. Quantitative analyses of the approach on simulated data, and qualitative analyses on genuine MR images of the foot and hip, demonstrate the superiority of the combined image approach over the equivalent algorithm applied to individual images. The technique has the potential for application to situations where a bias field correction is needed, but likely to be unstably constructed for the acquisition of interest.