Magnetic resonance imaging image intensity correction with extrapolation and adaptive smoothing
Hu Cheng, Feng Huang · Magnetic Resonance in Medicine · 2006
A significant problem in magnetic resonance imaging (MRI) is the inhomogeneity of the image resulting from a number of factors that are hardware related. The obtained image can be treated as the true image multiplied by a signal modulator, which is usually smooth across the image. A class of MR image intensity correction methods extracts the slowly varying component from the image with low-pass filtering or smoothing to approximate the signal modulator. This usually causes the edge enhancement artifact in the corrected image. A novel method of extrapolating the image in advance is proposed to reduce this effect significantly. Closest point algorithm is implemented to minimize the calculation time for extrapolation. To remove bright spots caused by nonuniform sensitivity profiles, a gradient-weighted smoothing method is discussed in this work. The partial differential equations based model is applied for locally adaptive smoothing. The filtered gradient of the corrupted image is used as the weight for smoothing. Phantom and clinical data collected on various MRI systems are used for evaluation of our method. These experimental results show that the proposed method solves the edge enhancement and bright spots problem effectively and robustly.