Referenceless multi-channel signal combination: A demonstration in chemical-shift-encoded water-fat imaging

Without explicit coil sensitivity information acquired by means of a reference body coil, multi-channel signal combination for water-fat separation (WFS) can be challenging due to channel-dependent phase offsets and chemical-shift dependent phase shifts. This study aims to develop a referenceless, robust, accurate, and fast channel combination method for WFS.

A dual-step multi-channel combination method is proposed. In the first step, channel-dependent phase offsets are estimated with a preliminary WFS estimation. In the second step, the multi-channel data are combined after removing phase offsets. Thereafter, WFS is performed to obtain final results. Numerical simulations (4-64 coils) and in vivo experiments (8, 16, 28 coils) at 3T field strength are conducted to compare the proposed method to previous methods. Channel combination with a body-coil scan serves as the reference for in vivo experiments.

The proposed method estimates channel-dependent phase offsets accurately. It shows improved robustness to phase singularities than weighted mean and adaptive reconstruction. It is faster than adaptive reconstruction (e.g., 25.45 versus 46.34 s with 28 coils) and the channel-by-channel WFS method (e.g., 21.77 versus 50.04 s with 8 coils). It provides comparable fat quantification accuracy to the reference under various reasonable signal-to-noise ratio conditions (e.g., Pearson correlation coefficient r = 0.981 with P < 0.01, for in vivo fat fractions using flip angle = 10°).

The proposed referenceless channel combination method may be beneficial to both qualitative and quantitative water-fat imaging.