Figure 1.
Flow chart of the multi-reception strategy.
Figure 2.
4-element phased array divided into two groups for 2 receive channel system.
Figure 3.
Graphic illustration of the extended GRAPPA.
Figure 4.
Hardware modification for multiple receptions.
Figure 5.
Photos and schematics of Nova 16-element and 8-element human brain receive-only arrays for 7T.
Both coil arrays have under-lapped elements in axial plane. 8-element array has one row, while 16-element array has two overlapped rows.
Figure 6.
SNRs of silicone oil phantom images of 16-element Nova array with various group partitions: a-d, combined images with 8 receive channels and 2 receptions; e, combined image with single reception and 16 receive channels.
It shows that non-overlapped and non-adjacent coils in the same group achieve higher SNR. The image SNR in (a) has 16% SNR gain than (e), while partition in (d) only provides a little SNR benefit.
Figure 7.
SNR maps derived from eight-element Nova array, sixteen-element Nova array with 16 receivers, sixteen-element Nova array with 8 receivers (N = 2, R = 2) and sixteen-element Nova array with 8 receivers (N = 2, R = 3) at (a) axial plane, (b) sagittal plane and (c) coronal plane.
Note that 1.15–1.25 SNR decrease for R = 2 and 1.30–1.50 SNR decrease for R = 3, lower than and
respectively. It is mainly due to the multiple receptions, additional ACS lines acquired for GRAPPA.
Figure 8.
Multiple receiption strategy with proper coil grouping shows the significant noise correcation (NC) improvements, which is helpful to understand the SNR enhancement with this method.
The noise correlation matrix (a), after re-arrangement with the optimized coil grouping, the largest values are relocated into the I and III quadrants (b). After dual receiption, those large values in I and III quadrants are then removed (c), and filled with 0 (d).
Figure 9.
Images, fitting errors and group partitions of 16-element Nova brain array with extended GRAPPA.
Accelerate rate R is 3. Group partition (a) and (b) which avoid overlapped coils in groups, can achieve less fitting error (<20%) than the regular GRAPPA in (e). Group partition (c) and (d) which have overlapped coils in groups, lead to more artifacts in images than that of the regular GRAPPA. 10 ACS lines are chosen, which is acquired inplace and used for the final images. Kernel size is 4, which is composed with two points above and two points below (similar to the method shown in reference 31).