Purpose: Signal magnitude can robustly be combined using the sum-of-squares approach. Methods have been developed to combine complex images. However, techniques based only on signal phase have not been developed and evaluated.
Theory and Methods: We performed simulations to demonstrate the effect of noise on coil combination. 32-channel 7 Tesla human gradient echo MRI brain data were collected. We combined phase images based on phase noise leading to spatially selective and coil selective combination of phase images. We compared our selective combination approach to optimal noise distribution and adaptive combination methods.
Results: We found that selective combination of signal phases leads to improved phase signal-to-noise ratio. Furthermore, a phase shift can be present in combined phase images introduced by the method used to combine multiple channel phases.
Conclusion: Mapping of signal phase from ultra-high field MRI data undoubtedly provides a wealth of information about the ageing brain and the effects of neurodegenerative disorders. Measurement of signal phase is essential in frequency shift mapping and in quantitative susceptibility mapping. The method used to combine signal phase should be informed by an understanding of the noise distribution in signal phase at the individual channel level.