Phase imaging benefits from strong susceptibility effects at very high field and the high SNR afforded by multi-channel coils. Combining the information from coils is not trivial, however, as the phase that originates in local field effects (the source of interesting contrast) is modified by the sensitivity profile of each coil. This has historically been addressed by referencing individual coil sensitivities to that of a volume coil – the Roemer/SENSE method¹ - but an alternative approach is required for ultra-high field systems in which no such coil is available. A method has recently been proposed - COMbining Phased array data using Offsets from a Short Echo-time Reference, or COMPOSER² – which approximates the phase measured in a reference scan with a short echo time to the phase of the receiver sensitivity and other phase effects present at TE=0. In this study we assess the effectiveness of COMPOSER in phase imaging with calf and breast coils in which the local transmit coil is not capable of receiving signal (i.e. where the Roemer approach cannot be applied) and quantify its performance with respect to another reference-coil-free method, MCPC-3D, and the Roemer method in the brain.

Measurements of 8 healthy subjects were made with a 7 Tesla MR whole body Siemens Magnetom scanner. High resolution T₂^*-weighted gradient-echo data of the brains of six male subjects were acquired with a 32 channel ¹H head coil (Nova Medical) (0.3x0.3x1.2 mm³, TE/TR = 15/28 ms, TA=12 mins), of the calf in one female subject with a two-channel ¹H coil⁴ (0.55×0.55×2.0 mm³, TE/TR = 10/15 ms, TA=101 secs) and of the breasts of one female subject with a four-channel dual-tuned ³¹P/¹H coil (Stark Contrast) (1.6×1.6×1.3 mm³, TE/TR = 3/7 ms, TA = 102 secs). The short echo reference (SER) data for COMPOSER were acquired with a 3D variable TE (vTE) sequence⁵ (2x2x4 mm³, vTE/TR = 0.8/5 ms, TA = 11 s). In the brain, dual-echo gradient-echo scans with 2x2x3 mm³ resolution were also acquired for the MCPC-3D method (TE/TR = (5, 9)/606 ms), and with both the birdcage transceive coil (AC) and the receive array (VC) for the Roemer method (TE/TR =5/606 ms).

The real and imaginary data from the SER scan were coregistered to the high resolution scan with FSL’s FLIRT⁶. A combined phase image was calculated from the single-channel magnitude (M_j) and phase images (θ_j) from each of the j coils of the high resolution data using

$$\theta_{COMPOSER}=\angle\sum_jM_j\cdot e^{-(\theta_j-\theta_{SER,j})}$$

where $$$\angle$$$ is the angle and θ_SER,j the phase of the SER scan in the space of M_j and θ_j. A modified Roemer reconstruction (with correction of phase only) was carried out according to $$\theta_{Roemer\_mod}=\angle\sum_jM_j\cdot e^{-(\theta_j-(\theta_{VC,j}-\theta_{AC}))}$$ where (θ_VC,j-θ_AC) is the phase correction, which was complex-smoothed and likewise in the space of M_j and θ_j.

The MCPC-3D-II reconstruction was carried out according to Ref. 3. The quality of phase matching in each voxel was assessed via the metric ‘Q’, where θ_cor,j is the corrected phase in the exponent in each method;

$$Q=100\times\frac{abs(\sum_jM_j\cdot e^{-\theta_{cor,j}})}{\sum_jM_j}$$

which approaches 100% for perfect matching.

The absolute value of the complex sum (‘Magnitude’) and Q values were low for No Correction (Figure 1, top two rows) (median Q over all subjects = 19.0%) and close to 100% for the reference Roemer method (median Q=99.2%). Magnitude and Q values were generally high with MCPC-3D-II (median Q=96.9 %), although errors in unwrapping single-channel phase values led to isolated low values (Figure 1, arrows 1-3). Phase matching with COMPOSER was similar to that with Roemer, with a median Q value of 98.9%. These observations were confirmed by the histogram analysis over all voxels and subjects in Figure 2. COMPOSER was similarly effective in the calf (Figure 3) and breast (Figure 4), yielding near-perfect phase matching throughout the image and artefact-free combined phase images which could be unwrapped.COMPOSER is a fast, robust method for the phase-sensitive combination of data from coil arrays. It requires no reference volume coil, making it feasible for use with all coil arrays, including PTx coils and surface arrays where the transmit part may not be engineered to receive signal. COMPOSER needs no phase unwrapping and provides phase matching which is comparable to the Roemer method and superior to that with the rival reference-coil-free approach tested.