Model-based intensity nonuniformity correction in brain MRI

Intensity nonuniformity is an artifact frequently observed on magnetic resonance (MR) images. It consists in anatomically irrelevant intensity variations that often lead automated segmentation algorithms to fail. This paper presents a novel approach for correcting those intensity variations. It is based on the assumption that clues about intensity nonuniformity can be found by modeling the acquisition sequence used. Thus, we develop a model from which transmitted and received B₁ fields can be computed for both homogeneous and heterogeneous objects. The proposed method has been evaluated on both physical phantom and human brain datasets using conventional quantitative criteria. The results show a significant reduction in intensity nonuniformity, especially for severe cases, and demonstrate a new wax of dealing with this type of artifacts.