Using the Phase for MR Image Segmentation: Application to the Bone of the Knee
Authors: Bourgeat, P., Stanwell, P., Ramadan, S., and Ourselin, S.
Journal: Proceedings of the 14th International Society of Magnetic Resonance in Medicine (ISMRM) Scientific Meeting
Abstract:
In MR, the signal is acquired in K-Space, resulting in a complex image composed of a phase and a magnitude. With a properly chosen pulse sequence, the phase can give information about tissue interfaces, but phase is only defined within the interval [0, 2π), and phase unwrapping is required prior to processing. Phase unwrapping is extremely time consuming, especially in 3D, and is prone to errors. As a result, only the magnitude of the complex MRI signal is used for clinical diagnostics, leading to a loss of information which is, and can only be, encoded in the phase of the signal. For example, a closer look at the articular and meniscus cartilages shows that the phase is more efficient in separating the two cartilages than magnitude, which is a common problem in cartilage segmentation. We present a novel method to extract textural information from the phase image without the need of phase unwrapping. The technique was applied for the segmentation of the bone of the knee as a basis for cartilage segmentation of the osteoarthritic knee. This work intends to show the potential benefit in including the phase information in MR segmentation algorithms.
Download the paper (PDF: 997KB)