Current rigid transmit-receive RF coils for MR imaging of the knee do not adapt to the circumference of the human knee, which varies significantly with body weight. If the knee is small and the coil is poorly loaded, coil losses can become significant. In the case of patients who are too large to fit the knee coil, flexible surface coils are used, which do not allow optimal placement and configuration of every coil element.

Therefore, a new coil design which adapts to the individual size of the knee is desirable. This work analyses such a concept for 1.5T, proposing a design that enables the array to adjust by keeping the element size constant but allowing the overlap between adjacent coil elements to change. SNR simulations and measurements of the coil properties were performed for three different phantom sizes, corresponding to three design configurations. We investigated the tradeoff between maintaining optimal coil loading and SNR losses due to strong coupling. To mitigate coupling effects, broadband matching was applied [1].

Coil array designs were considered with array radii from 75mm to 110mm, arranged in three rows of eight fixed-sized coil elements equally distributed around a cylinder. The overlap between the coil elements decreases with increasing phantom size (Figure 1). As coupling is highly dependent on the degree of overlap [2], the dimensions of the elements were chosen such, that the coupling coefficient $$$k=\Delta f/f_{R}$$$ for the maximal overlap does not exceed the k of touching coil elements (configuration b: center of each element with width d is separated by $$$l=1.0d$$$).

The SNR of the array was computed for the three phantom sizes shown in Figure 1, using a full-wave electrodynamic simulation tool based on dyadic Green’s functions (DGF) [3]. The DGF approach takes coil losses and sample noise into account, but does not include inductive coupling. The expected degradation due to coupling was determined by deriving the product of k and Q from bench measurements with two adjacent coil elements, constructed to match the simulated coils.

To mitigate coupling, we applied broadband matching instead of a classic match with $$$Z_{opt}=50\Omega$$$. Therefore, the coil impedance was matched to $$$Z_{opt}=50\Omega\cdot|1\pm jkQ|$$$ [1]. Figure 2 shows the SNR degradation for all degrees of mismatch. The reflection coefficient which provides the best SNR results over the whole range of kQ was chosen as broadband match.

The performance of the adaptive coil was compared with a simulation of a standard 24-channel knee coil with a fixed radius of 86.5mm. Nearest neighbors are assumed to be decoupled, therefore a classic match was applied, and only SNR degradation due to next-nearest neighbors coupling was determined. The expected SNR was evaluated for the small- and middle-sized phantom, which represents the maximal phantom size for this coil.

Measurement and simulated SNR results are shown in Figure 3. As can be seen, the Q-Factor of the standard coil increases about 50% from the middle- to the small-sized phantom, whereas it is almost independent of the phantom size when the coil adjusts to the circumference of the phantom. Due to lower coil losses with small phantoms, the adaptive coil can reach higher SNR than the standard coil when coupling effects are not considered. For the big phantom, SNR decreases as sample noise increases with phantom size.

Nevertheless, there is significant coil coupling for the adaptive coil (Figure 4). For touching coil elements (adaptive middle), kQ reaches its maximum of 7.08, which leads with the preamp noise figure $$$NF_{min}=0.6$$$ to an SNR degradation of 50%. A broadband match for $$$kQ=7.08$$$ with $$$s_{11}=0.75$$$ can increase the expected SNR to 75% for the middle-sized coil and to 80% for the small coil.

However, the non-ideal geometric overlap of this design has a higher impact on SNR degradation than the additional coil losses associated with a non-perfectly fitting coil. By adjusting the broadband match for every configuration, the SNR could individually be optimized, particularly for configurations whereby kQ differs significantly, but this would be complicated to implement. To compensate the strong coupling of this concept sufficiently, it would be necessary to reduce NF_min.

Our results show, that an adaptive coil can achieve lower loaded coil losses than standard coil designs and therefore could be implemented to expand the array size without dramatic SNR losses. Nevertheless, it is necessary to mitigate the SNR degradation due to inductive coupling to improve overall performance compared to standard coil designs. One potential way is to keep the coil elements decoupled by adjusting their dimensions to maintain the optimal overlap ratio [4].