Receive array performance is generally maximized when the array conforms closely to the body. People vary in size, making it difficult for one coil array to fit optimally on all subjects. This is particularly true for rigid coils, such as knee coils, which must be made large enough to accommodate most people, but which will not be optimal on smaller bodies. With the increasing prevalence of obesity, there are many people who do not even fit in the standard commercially available knee coils, and must be imaged with improvised arrangements of body array coils.

Several coil array designs have been reported which adapt to the size of the object being imaged, either by allowing fixed-size coil elements to move¹ or by physically stretching coil elements made from copper braid². Both of these approaches lead to changes in coil tuning, match and decoupling such that the optimum noise-matched impedance is not always presented to the preamp. A size-adjustable stripline head array has been presented which addresses some of these issues with a mechanically driven decoupling mechanism³. We present here a cylindrical array in which surface-coil elements of fixed circumference morph as the array is expanded such that there is little change in tuning, match or decoupling. An underlying trellis-like framework controls the coil element configuration. The prototype 3T array is assessed for Q, S-parameters and SNR on 3 different-sized phantoms corresponding to a range of human knee dimensions.

The underlying framework is a lattice of 5mm wide strips of 2.4mm thick nylon. Holes were drilled every 5.7mm and the strips cut out using a circuit board router. Two layers of 72 strips are linked together with nylon screws and nuts to form a lattice (Fig.1). Changing the angle between the strips in the two layers allows the lattice to be stretched in one direction or the other. The lattice is wrapped around to form a cylinder, and circuit-board components are mounted on the lattice with a hole and slots such that the boards ride on the lattice as it stretches (Fig.2). More detail of the design principles are given elsewhere [4]. Multi-strand Teflon-coated wire is used to link the circuit board elements and is pulled around screws located at the corners of the coil element to define the coil element shape. The use of 10 nylon strips per coil element width allows a 10% coil overlap for inductive decoupling. The lattice can expand in diameter from 14.5cm to 20.5cm, morphing the coils from 6.5×9.5cm to 9.5×6.5cm in the process (Fig.3). Each element is tuned to 123.2 MHz with 4 capacitors and has an active detuning trap. The coils were interfaced to the scanner through an in-house built 8 channel interface. The array was evaluated on three cylindrical phantoms with ε_r=56.6, σ=0.37 S/m with diameters of 13.4, 16.5 and 19.5 cm. It was also compared to a commercially available 15 channel knee coil (QED, Mayfield Village, Ohio).The unloaded-to-loaded Q ratio was 230/35 for the large phantom, dropping to 230/50 on the small phantom. A conventional circuit-board coil element of the same size had an unloaded Q of 337. The elements of the 8 channel array were tuned and matched, and decoupling was adjusted by moving a wire jumper on the circuit boards. Once this was set no adjustments were made to the array as it was placed on the different sized phantoms. S-parameter plots for a pair of coils are shown in Fig.4. Coil tuning, match and decoupling all remain within acceptable limits as the array is stretched and the coil elements morph to different shapes. Worst case S₁₁ and S₂₁ are -12dB and -10.7dB respectively. SNR plots for sum-of-squares reconstruction are shown in Fig.5. Central SNR increases from 22 to 29.9 to 42.2 as the phantom size is decreased. For comparison, if the coil remains fixed at the largest diameter, SNR in the smallest phantom is only 33.4 (Figure 5, fourth column). The commercial coil, which has a fixed internal diameter of 16cm, provided SNR of 33 in the small phantom. The other phantoms would not fit in it.The coronal SNR maps show how the z-extent of the trellis array shrinks as the array diameter increases. Central SNR for the large phantom would be improved if the array was longer, which could be achieved by adding extra rows. This would also help match the larger z field of view of the QED coil. This design may have many applications including pediatric coils, head coils and body arrays^5.