An orthogonal correlation algorithm for ghost reduction in MRI.
Ghosting in MRI due to modulation of k-space data can be caused by motion of the subject or characteristics of the sequence. A general solution for 2DFT MRI that reduces ghosting without causal modeling is presented. Separate image data sets are acquired in which the phase and frequency directions are swapped. In these two data sets, the image signal is correlated, whereas the ghost signals are not. By taking a correlation of these two data sets, an image with greatly reduced ghosting is obtained. The reduction is shown to depend both on the correct signal intensity of the image, as well as the ghost intensity in the ghosted region. The reduction approaches 100% in regions of low image signal, and is more moderate in regions of higher image signal. The process was applied to conventional spin-echo, fast-spin-echo, and gradient echo imaging of volunteers and a phantom. Results of a reader study of the volunteer images reflected a significant overall reduction of ghosting artifacts in all volunteer experiments.