Two-Step Generic Referenceless Phase Combination (GRPC) for accurate phase image reconstruction from multiple receiver coils
Francesco Santini1,2, Mathieu D Santin3, Paulo Loureiro de Sousa4, and Oliver Bieri1,2

1Radiological Physics, University of Basel Hospital, Basel, Switzerland, 2Department of Biomedical Engineering, University of Basel, Basel, Switzerland, 3Institut du Cerveau et de la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France, 4CNRS, ICube Laboratory, FMTS, Université de Strasbourg, Strasbourg, France

Synopsis

This work presents a method to combine the signal from multiple coils in order to obtain a coherent phase image. The methods is agnostic to the acquisition protocol and the coil geometry, and does not require operator interaction.

Purpose

Phase imaging is especially important at high to ultra high field strengths, both to map local field inhomogeneities and to highlight important anatomical structures (e.g. with Susceptibility Weighted Imaging, SWI). Although nowadays multiple coils are commonly used for MR image acquisition, accurate phase combination is not yet generally offered by the generic reconstruction pipeline of commercial scanners. The reason is that the various receiver coils experience different phase sensitivities due to their geometry, and the complex data need to be manipulated before combination in order to avoid destructive interferences. Last year, a generic framework was presented by Santini et al.1 that allowed phase reconstruction for arbitrary sequences without any user interaction or prior knowledge, termed Generic Referenceless Phase Combination (GRPC). While the solution was optimal in a mathematical sense, the most robust embodiment was implementing a constant phase correction, therefore not taking into account spatial variations. In this work, we use the zeroth-order GRPC method as a first step of the correction, and then apply an additional correction as proposed by Parker et al.2. The method is demonstrated in the brain and in the abdomen using two different acquisition sequences and the quality of the reconstruction of the two steps is compared by using a quality of phase matching metric3.

Methods

The two-step GRPC correction was implemented as a generic functor of the reconstruction pipeline of a commercial 3T scanner by separately performing two correction step. The first step was a zeroth-order correction (removal of average coil phase) for preliminary coil rephasing. Subsequently, the rephased images were processed through a modified Parker virtual coil approach2 by filtering the difference between the true coils and the rephased complex sum with a 21x21 2D gaussian low-pass filter. The reconstruction was applied to two different sequences, a gradient echo and a balanced steady-state free precession. Images were acquired on a healthy volunteer in the abdomen with a 18-channel body array together with an integrated spine coil (matrix size 256x154x1, resolution 1.6x1.6x5mm3, GRE TR/TE 15/3.6ms, bSSFP TR/TE 4.2/2.1ms), and in the head with 20- and a 64-channel coils (matrix size 192x192x1, resolution 1.0x1.0x5mm3, GRE TR/TE 45/4.7ms, bSSFP TR/TE 4.6/2.3ms). In the reconstructed complex images, the quality of phase matching (Q) map was calculated pixelwise according to the formula: $$Q(x,y)=\frac{||\sum I_c(x,y)||}{\sum||I_c(x,y)||}$$, where Ic is the complex signal intensity detected by each coil c. The median value of Q was calculated on each map (only in the portion of the images where the signal level is above the noise).

Results

The correction produced artifact-free phase images for all location and coils (Fig. 1). The phase images produced by GRPC and 2-step correction appear similar, however a lower noise level is recognizable in the two-pass images (Fig. 2). Quantitatively, the Quality of phase matching maps (Q-maps) showed important differences between the two methods (Fig. 3). The median Q values across the different coils and position averaged 0.56±0.19 for GRPC and 0.95±0.06 for the two-step correction (Fig. 4).

Discussion

For the combination of multi-element coils, both the GRPC and the proposed method deliver consistent phase images. However, the addition of the second step greatly increases the phase matching with which the coil elements sum, meaning that the coil sum constructively over the whole field of view. This increases the signal level in the combined image and has a noticeable impact on the signal-to-noise of the phase images. This method was used to implement a very generic reconstruction functor, which could be applied to any combination of protocol parameters and delivered fast reconstruction seamlessly at the scanner console, without need for external postprocessing. We showed that the method works equally well, and without operator interaction, with different coils and type of acquisition sequences.

Conclusion

The addition of a second processing step to the GRPC method is beneficial to delivering accurate phase images from multi-element coils.

Acknowledgements

No acknowledgement found.

References

1. Santini F, Ganter C, Ehses Ph, et al., A Generic Referenceless Phase Combination (GRPC) Method: Application at High and Ultra-High Fields, Proceedings of the ISMRM 2015

2. Parker DL, Payne A, Todd N, Hadley JR. Phase reconstruction from multiple coil data using a virtual reference coil. Magn. Reson. Med. 2014;72:563–569

3. Robinson SD, Bogner W, Dymerska B, et al, COMbining Phased Array Data Using Offsets from a Short Echo-Time Reference Scan (COMPOSER), Proceedings of the ISMRM 2015

Figures

Phase maps delivered by simple complex sum and the two investigated methods. Black arrows point to phase wraps and red arrows point to artifacts generated by the coil combination (physically-impossible wraps).

Detail of a phase image of the liver reconstructed with GRPC (A) and the two-step combination (B). A lower noise level with the two-step method is especially noticeable in the proximity of the phase wrap (pointed by the arrow).

Quality of Phase Matching (Q) maps for the two investigated methods. A value of 1 corresponds to perfect phase matching among the coils.

Box plot of the median Q values at different location/coils compared between the two investigated methods.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
4314