Typically, coil elements or arrays are dispersed on a two-dimensional surface to ensure their sensitivity profiles do not overlap, since correlated noise mitigates an SNR improvement when overlapping coils are operating in the sample-noise-dominated regime. In this study, we show that a small flexible array, operating in the coil-noise-dominated regime, can locally improve temporal SNR when placed inside a whole-head array. The two concentric arrays are inductively decoupled using preamplifier decoupling, and the contribution of coil noise to the overall noise reduces the noise correlation. Up to a two-fold increase in temporal SNR is achieved in the motor cortex.
A 7-channel, flexible receive coil was constructed that could be placed inside a previously published whole-head, 24-channel receive coil for non-human primates5 (Figure 1). Each element of the flex coil was 2.5-cm wide, producing a total array width of 6.4 cm. Each element was connected to a low-input-impedance preamplifier by a coaxial cable with a half-lambda electrical length (including two lattice baluns, a phase shifter, and a high-density coaxial connector). All coils were tuned to 297 MHz.
The Q-ratios of the flex coil were measured for an isolated element and when elements were placed in the completed array with coaxial cables and lattice baluns attached. The Q-ratio of a representative element of the whole-head coil was likewise measured. The S12 between elements of the flex coil and between a representative element of the whole-head array was measured directly from the preamplifier input ports. Preamplifier decoupling—essential for reducing inductive coupling (and therefore noise correlation) between the two concentric arrays—was measured using a standard double-probe technique6.
To evaluate temporal SNR, the flex coil was placed over the motor cortex of a rhesus macaque. The whole-head coil was placed directly over top of the flex coil. The thin flex coil required only ~2.5 mm of radial space. Multi-band EPI7 time-series were acquired with both the whole-head coil and flex coil active and subsequently with just the flex coil active. The temporal SNR was calculated from the de-trended time series on a pixel-by-pixel basis. All imaging was performed on a 7T head-only scanner.
The Q-ratio of an isolated flex element was 4.8. When placed in the completed array with associated cables and baluns, the Q-ratio was reduced to 1.2 – 1.4; this indicated the flex coil was operating in the coil-noise-dominated regime and coil noise accounted for 71% to 83% of the total noise. The Q-ratio of the representative whole-head coil element was 3.4, indicating that coil noise contributed only 29% of the total noise.
Without preamplifier decoupling, the mean S12 between elements of the flex coil was -22 dB. The worst-case isolation between an element of the flex coil and an element of the whole-head coil was approximately -8 dB; however, preamplifier decoupling provided an additional -20 dB of isolation between elements of the two arrays. This resulted in a mean noise correlation between elements of the flex coil of 18% and an overall correlation of 10% for all elements. The maximum noise correlation between any element of the whole-head coil and any element of the flex coil was 23% (Figure 2).
The ratio of the temporal SNR when the whole-head coil and flex coil were active versus when only the whole-head coil was active is provided in Figure 3. The temporal SNR in the motor cortex improved by up to a factor of two and improvements in temporal SNR were still realized deep within the brain.
Preamplifier decoupling sufficiently reduced through-space inductive coupling between the flex coil and the whole-head coil. Through-sample, resistive coupling is unavoidable; however, the contribution of coil noise to the overall noise allowed the flex coil to remain largely uncorrelated with the surrounding whole-head coil. The result was a significant increase in temporal SNR, despite the sensitivity of the two arrays being entirely overlapped.
Traditionally, coil arrays are arranged on a two-dimensional surface to ensure their sensitivity profiles do not overlap, since there is diminished SNR benefit when coils are operating in the sample-noise-dominated regime. We have shown that a significant improvement in temporal SNR can be achieved by adding a local coil array that operates in the coil-noise dominated regime. This topology becomes increasingly useful when array elements become small, such as when imaging rodents and non-human primates.