A dual image approach for bias field correction in magnetic resonance imaging Shang-Hong Lai a, *, Ming Fang b a Department of Computer Science, National Tsing-Hua University, Hsinchu, Taiwan 300 b Siemens Corporate Research, 755 College Road East, Princeton, NJ 08540, USA Received 2 September 2000; accepted 19 October 2002 Abstract In this paper, we propose a dual image approach to correcting intensity inhomogeneities for MR images acquired using surface coils. Previous methods are usually not satisfactory due to restricted application domains, considerable human interactions, or some undesirable artifacts. The proposed algorithm provides nice correction results for a variety of surface-coil MR images. It is accomplished by using an additional body-coil MR image of a smaller size captured at the same position as that of the surface-coil image to facilitate the estimation of the bias field function. The correction algorithm consists of aligning the surface-coil image with the body-coil image and fitting a spline surface from a sparse set of data points for the associated bias field function. Experiments on some real images show satisfactory correction results by using the proposed algorithm. © 2003 Elsevier Science Inc. All rights reserved. Keywords: Intensity inhomogeneity correction; Medical image restoration; Image fusion; Spline surface fitting; Magnetic resonance imaging 1. Introduction Spatial intensity inhomogeneity induced by the radio frequency (RF) coil in magnetic resonance imaging (MRI) is a major problem in the computer analysis of MRI data. For example, the spatial intensity inhomogeneity has made the classification task in MRI very difficult for most of the existing segmentation methods [1]. This is due to the fact that the intensity inhomogeneities appeared in MR images produce spatial changes in tissue statistics, i.e., mean and variance. In addition, the degradation on the images ob- structs the physician’s diagnoses because the physician has to ignore the inhomogeneity artifact in the corrupted im- ages. The spatial inhomogeneity is particularly severe in MR images acquired using surface coils. However, the surface-coil MR images can provide better signal-to-noise ratios than the body-coil images. In this paper, we concen- trate on the problem of intensity inhomogeneity correction for surface-coil MR images. The correction of spatial intensity inhomogeneity has been regarded as an essential post-processing step to the computer analysis or processing of MRI. The removal of the spatial intensity inhomogeneity from MR images is difficult since the inhomogeneities could change with different MRI acquisition parameters, from patients to patients and from slices to slices. Therefore, the correction of intensity inho- mogeneities is usually required for each new image. A number of methods have been proposed for the inten- sity inhomogeneity correction, or bias-field correction, problem in the last decade. They can be roughly classified into three approaches namely, the phantom-based approach [2,3], the homomorphic filtering approach [1], and the sur- face-fitting approach [4 –7]. Unfortunately, these ap- proaches usually either work under restrictive conditions, require considerable user interactions, or render undesirable artifacts. In this paper, we propose an inhomogeneity surface fit- ting technique for correcting the surface-coil MR image by utilizing an additional body-coil MR image of a small size acquired at the same position as that acquired for the sur- face-coil image. The fusion of the surface-coil image with the additional body-coil image facilitates the estimation of the inhomogeneity surface due to the fact that the body-coil images in general are much more homogeneous in the intensity distributions for the same tissue regions than the surface-coil images. The two images are first aligned * Corresponding author. Tel.: +886-3-574-2958; fax: +886-3-572- 3694. E-mail address: lai@cs.nthu.edu.tw (S.-H. Lai). Magnetic Resonance Imaging 21 (2003) 121–125 0730-725X/03/$ – see front matter © 2003 Elsevier Science Inc. All rights reserved. S0730-725X(02)00637-9