SEGUE: a Speedy rEgion-Growing algorithm for Unwrapping Estimated phase

Anita Karsa, Karin Shmueli
2019 IEEE Transactions on Medical Imaging  
Recent magnetic resonance imaging (MRI) techniques, such as quantitative magnetic susceptibility mapping, employ the signal phase to reveal disease-related changes in tissue composition, including iron or calcium content. The MRI phase is also routinely used in functional and diffusion MRI for distortion correction. However, phase images are wrapped into a range of 2π radians. Phase region expanding labeller for unwrapping discrete estimates (PRELUDE) is the gold standard method for robust,
more » ... ial, 3-D, MRI phase unwrapping. Unfortunately, PRELUDE's computation time can reach 15 min for a severely wrapped brain image and nearly 10 h to unwrap a full head-and-neck image on a standard PC. In this paper, we develop a Speedy rEgion-Growing algorithm for Unwrapping Estimated phase (SEGUE) based on similar principles to PRELUDE, implemented with additional methods for acceleration. We compared PRELUDE and SEGUE in numerical phantoms, and using in vivo images of the brain, head and neck, and pelvis acquired in 4-5 healthy volunteers and at 4-6 echo times. To overcome chemical-shift-induced errors within the head and neck, and pelvic images, we also investigated applying both techniques within fat and water masks separately. SEGUE provided almost identical unwrapped phase maps to the gold standard PRELUDE. SEGUE was (1.5 to 70 times) faster than PRELUDE, especially in severely wrapped images at later echoes and in the head and neck, and pelvic images. Applying these techniques within fat and water masks separately removed chemical-shift-induced errors successfully. SEGUE's MATLAB implementation is available for download. SEGUE is a general unwrapping algorithm not specific to MRI, and therefore could be used in images acquired with other modalities.
doi:10.1109/tmi.2018.2884093 pmid:30561341 fatcat:gdrjbzwwn5f4va2hbzb43ejzly