### 4503

Dynamic Decoupling for Simultaneous Transmission and Acquisition in MRI
Bilal Tasdelen1,2, Alireza Sadeghi-Tarakameh1,2, Ugur Yilmaz2, and Ergin Atalar1,2

1Department of Electrical and Electronics Engineering, Bilkent University, Ankara, Turkey, 2National Magnetic Resonance Research Center (UMRAM), Ankara, Turkey

### Synopsis

In order to use simultaneous transmission and acquisition in clinical MRI for living subjects, robustness to load and environmental changes has to be established, especially for uncooperative subjects. High isolation can be achieved with active cancellation methods, but maintaining it over a long time is a challenge. A look-up table based method is proposed with a smart search algorithm that enables fast dynamic decoupling of transmit/receive coils using an active decoupling circuit. Experiments with a birdcage coil used as a transceiver show that maintaining >80 dB isolation is possible even under the presence of load variation.

### Introduction

In conventional MRI, excitation and acquisition is interleaved due to coupling of transmit signal to MR signal. Simultaneous transmission and reception in MRI requires high isolation (>100 dB) due to small MR signal amplitude, but it has the benefits such as acquiring signal from short T2 tissues and less RF power consumption.1 Although there are active and passive decoupling methods proposed2-6 in order to get rid of the leakage signal, shortcomings of these methods hinder the possibility of clinical usage. Sensitivity of decoupling to load variation and environmental change is one such challenge.3 There are optimization-based active decoupling methods proposed using a self-interference cancellation circuit, yet they are not suitable for dynamic decoupling where abrupt changes of coil parameters may occur.4,5

In this study, a novel algorithm is proposed where STAR (Simultaneous Transmit and Receive) circuit parameters are controlled to counteract environmental and load changes that deteriorate decoupling performance over time.

### p { margin-bottom: 0.1in; line-height: 115%; background: transparent none repeat scroll 0% 0%; }a:link { color: rgb(0, 0, 128); text-decoration: underline; }a:visited { color: rgb(128, 0, 0); text-decoration: underline; }Methods

With the implemented STAR circuit topology (Fig.1), $m$ delay-attenuation lines with $n$ bit attenuators can generate $2^{nxm}$ possible vectors. Through the optimization of attenuation control parameters, a candidate vector can be found which cancels out the leak signal.4,5 However, there is no guarantee that problem will converge to global optimum, since the problem has many local minima.

Proposed algorithm can find one such candidate vector without doing any optimization by utilizing an offline look-up table. Therefore, without optimization process, high isolation can be achieved quickly and sustained even in the presence of load variation.

Considering the whole set of vectors that can be generated being finite, it is possible to store them, assuming the system is time invariant. Instead of measuring them all, which is not feasible due to time constraint, every possible output of the circuit can be modeled as superposition of signals coming from each line separately and stored beforehand. However, it is still time consuming to search in this offline stored look-up table due to its size. Proposed algorithm speeds up search process by sorting and sectioning this look-up table in a smart way (Fig. 2).

As result of this operation, first, a list is generated, where each element consists of a complex number (output of the circuit: Amplitude and phase) and a control input vector which will generate that output. Second, a matrix is generated that stores region boundaries that is determined by phase and amplitude intervals. When a new measurement acquired, the algorithm can point out the indices that include global optimum in the look-up table directly from this matrix. Using these outputs, leak signal can be estimated and cancelled out in a loop, correcting any bias in the characterization measurements.

### Results

Simulation performance of implemented STAR topology can be seen in Fig 3.

Using a network analyzer in place of MR scanner, decoupling experiments are conducted. A two port birdcage coil with a diameter of 20 cm is used as a transmit/receive coil. Experiment setup is shown in detail in Fig. 4. Circuit characterization is done when everything, including coils, are connected to circuit in order to demonstrate algorithm’s ability of eliminating bias coming from look-up table. Using these measurements, a look-up table is constructed with intervals of $Δφ=0.25°$, and $Δ|S_{21}|=10^{-5}$. Algorithm performance is tested over a time period under both stable and unstable environment. It can be seen that abrupt changes of load can be compensated in a few iterations (Fig. 5).

### Discussion

Ability to decouple the coils in a reactive manner makes it feasible to image living subjects. With this system, even if load changes slightly, it is possible to decouple coils dynamically and continue the scan, which can sustain high decoupling without sacrificing too much scan time.

One of the main drawbacks of this approach is high storage space requirement due to large look-up table. However, as storage space is getting cheaper, this drawback proves to be unimportant. Another drawback is very narrow band for decoupling, which can be solved by using a band pass filter as suggested by Sohn.3

### Conclusion

Sensitivity to load change of STAR in MRI is solved in this abstract by a novel algorithm that requires less iterations than optimization based algorithms to reach to the maximum decoupling condition, by using a look-up table together with an indexing like approach to search. By increasing response time to load changes, algorithm can increase the quality of imaging for living subjects, especially for uncooperative targets, which is of great concern in clinical imaging. Future usability of such circuit topology in conjunction with this algorithm for MRI will be demonstrated.

### References

1. D. J. Tyler, M. D. Robson, R. M. Henkelmann, I. R. Young, and G. M. Bydder, “Magnetic resonance imaging with ultrashort TE (UTE) PULSE sequences: Technical considerations,” J. Magn. Reson. Imaging, vol. 25, no. 2, pp. 279-289, Feb. 2007.

2. A. C. Özen, M. Bock, and E. Atalar, “Active decoupling of RF coils using a transmit array system,” Magnetic Resonance Materials in Physics, Biology and Medicine, vol. 28, no. 6, pp. 565–576, Dec. 2015.

3. S.-M. Sohn, J. T. Vaughan, R. L. Lagore, M. Garwood, and D. Idiyatullin, “In vivo MR imaging with simultaneous RF transmission and reception,” Magn Reson Med, vol. 76, no. 6, pp. 1932–1938, 2016.

4. Salim, Maryam, et al. "Detection of MR Signal during RF Excitation using Full-Duplex Radio System." Proc. Intl. Soc. Mag. Reson. Med. Vol. 24. 2016.

5. D. Bharadia, E. McMilin, and S. Katti, “Full duplex radios,” in ACM SIGCOMM Computer Communication Review, 2013, vol. 43, pp. 375–386.

6. A. C. Özen, E. Atalar, J. G. Korvink, and M. Bock, “In vivo MRI with Concurrent Excitation and Acquisition using Automated Active Analog Cancellation,” p. 3.

7. Y. Okada, T. Kawai, and A. Enokihara, “Design method of unequal Wilkinson power divider using LC-ladder circuits for multi-way power dividers,” in 2016 Asia-Pacific Microwave Conference (APMC), 2016, pp. 1–4.

### Figures

Fig. 1: Diagram of STAR circuit. A vector modulator consists of 4 delay-attenuation lines, each of them adds a different fixed phase, as well as attenuates the phase shifted signal by a certain amount that can be digitally controlled using serial communication. Power is divided/combined by a Wilkinson divider topology proposed by Okada7. This topology is chosen due to its flatter amplitude and phase response

Fig. 2: Demonstration of the algorithm for circuit characterization. $(φ_{ij}, |S_{21}|_{ij})$ is measured for every attenuation line i and state j. Look-up table constructed from superposition of possible states of lines contains signal constellation of circuit. After sorting $2^{mxn}$ complex numbers by their phases, resulting complex array is divided into regions by $Δφ$ phase intervals as shown in the orange region. Each such region is separately sorted by the amplitude. Boundaries of regions within a $Δ|S_{21}|$ interval are calculated as shown in the yellow region and stored.

Fig. 3: AWR simulation of STAR circuit. In the simulation, a simulated birdcage coil is considered as a transceiver antenna. As shown in the figure, using this topology, it is possible to achieve high isolation within a narrow band. As previously discussed3 and shown in the simulations in this study as well, narrow band problem is caused by discrepancy between S-parameters of coil and STAR circuit.

Fig. 4: Picture of the experiment setup. Network analyzer is controlled via Ethernet communication. To be able to control attenuators via SPI interface, a Raspberry Pi is used. STAR circuit is implemented in a modular fashion to be able to change circuit parameters on demand. Sampling period of the network analyzer is set to 250 ms. A birdcage structure is used as a transmit/receive coil which has 23 dB isolation between its ports.

Fig. 5: Experiments for decoupling of a birdcage coil under different circumstances. Upper figures show experiments when the environment is relatively stable. Bottom figures show the response of the system under load variation in the presence of body movement. Red regions at bottom figures show the times where additional load is moving. -80 dB is put as a threshold to pause the algorithm.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
4503