An Improved Region Growing Algorithm for Phase Correction in MRI

Purpose: To develop an improved region-growing algorithm for phase correction in MRI. Methods: Phase correction in MRI can sometimes be formulated as selecting a vector for each pixel of an image from two candidate vectors so that the orientation of the output is spatially smooth. Existing algorithms may run into difficulty in the presence of high noise, artifacts, or spatially isolated objects. In this study, we developed an improved region-growing algorithm to include the following novel and salient features: 1) automated quality guidance for determining the sequence of region growing, 2) joint consideration of two candidate vectors in selecting the output vector, and 3) automated segmentation during region growing for handling spatially isolated objects. The phase correction algorithm was tested in different body parts of five healthy volunteers at 3.0T and of one healthy volunteer at 1.5T for two-point Dixon water and fat imaging with flexible echo times. Results: The algorithm achieved successful phase correction in all the data sets tested, providing improvement in areas of known difficulty, when compared with an algorithm lacking the new features. Conclusion: The improved region-growing algorithm can be used for reliable and robust phase correction even when regions of high noise, artifacts, or spatially isolated objects are present in an image. (C) 2015 Wiley Periodicals, Inc.