SWIM has a built-in RF continuous wave (CW) signal generation for self-calibration, which makes it completely standalone. SWIM uses the magnitude of reflected power from the coil as feedback to compensate for the impedance mismatch. We also developed a pseudo-manual control of the tuning and matching network to adjust the impedance, irrespective of the automatic feature. Apart from automatic impedance matching function, this pseudo-manual control allows the user to fine-tune the impedance condition after the automatic function. The complete system is wireless and controlled by an in-house developed Android mobile application via Bluetooth.
RF coil and phantom design. A single loop RF coil was designed and fabricated. A housing structure was resin printed to securely hold the phantom close to the coil. The loop was chosen to have a diameter of 28 mm with a single distributive capacitor (1111C Non-magnetic capacitors, Passive Plus Inc, USA) to ensure uniform current distribution. An L-matching network with shunt-tune, series-match capacitors (SG9128, Sprague Goodman, USA) was chosen as shown in Fig. 2. To verify the performance of the coil, a centrifuge tube of 30 mm diameter was chosen as the phantom to represent a rat head. The tube was filled with distilled water, CuSO4 (1 g/L), and Agar (10 g\L). The phantom is Agar-based to mimic brain tissue properties and the combination of copper ions and Agar concentration affects the proton relaxation time constants. It was observed that copper ions predominantly control the T2 whereas, the T1 depends more on agarose concentration [38]. Phantom images were used to observe the change in the B1 field strength of the surface coil before and after the automatic calibration with the SWIM system.
The SWIM system. Manual tuning and matching of RF coils requires substantial time. Most coil systems present themselves with a sub-optimal matching condition after loading the sample. The loading effect creates an impedance mismatch and loss of signal that cannot be recovered by a pre-amplifier, thus deteriorating the SNR. The SWIM system is a fast, efficient, and cost-effective solution to impedance match any small animal 7T receive coils. It is an appropriate combination of automation and user input to obtain desired results. The concept behind the SWIM system is to compensate for the impedance mismatch caused by loading effect of a sample by cycling through multiple capacitor combinations and choosing the optimal impedance based on a least reflected power measurement. This system is completely standalone and automatic since it generates its own CW signal by employing a voltage-controlled oscillator (VCO) for the process and does not depend on RF gating signal from the console. An external gating signal from the console is used as an input to the system for software detuning of the receive coil during the transmit cycle and not to synchronize the system with console for automatic impedance calibration. The RF coil developed does not use any independent circuitry to detune, instead, software detuning uses one of the 256 combinations as a detuned state.
Self-calibration system. This module consists of a VCO (CVC055CW-250-450, Crystek Crystal Corp., USA) which can generate a CW RF signal with a power of -3 dBm. Two digital-to-analog converters (DAC) (MCP4725, Microchip Technology, USA) were used to adjust necessary voltages for the VCO. The microcontroller used inter-integrated circuit (I2C) serial protocol to interact with the DACs, which set up the supply and tuning voltages for the VCO. The RF signal was then followed by an E-pHEMT (Enhancement mode High-Electron-Mobility-Transistor) low noise amplifier (PHA-13LN+, Minicircuits, USA) to amplify the signal by 20 dB. A narrow band pass filter, designed in-house, was used as the next stage to filter out low frequency noise and higher harmonics from the VCO. The lumped element filter was designed to have a 54 MHz bandwidth at center frequency of 300 MHz. A Single Pole Double Throw (SPDT) CMOS RF switch (HSW2-272VHDR+, Minicircuits, USA) was used to switch the receive coil between MRI console (connecting to the receive chain LNA) and self-calibration module. RF common (RFC) port is connected to a bidirectional coupler, RF1 (port1) is connected to the MRI console and RF2 (port2) is connected to the self-calibration module. This switch offers high linearity with +85dBm 3rd order intercept point, low insertion loss, and internal CMOS driver, making it a perfect fit for low power impedance matching system. Insertion loss and isolation data of the switch were characterized at 300 MHz as 0.16 dB (RFC-RF1), 0.27 dB (RFC-RF2), and 47 dB (RFC-RF1/RFC-RF2) respectively. After the SWIM function, the switch will toggle to a slightly lower insertion loss path and the input impedance presented to the console can be fine-tuned wirelessly with the pseudo-manual function, via the mobile application.
Power measurement setup. A bi-directional coupler (ADCB-20-82+, Minicircuits, USA) is used to tap the reflected power. This device has a low measured insertion loss of 0.2 dB, 20 dB of coupling at both ports, and a directivity of 30 dB at 300 MHz. By terminating the forward coupling port with a 50 Ohm resistor, an accurate reflected power measurement can be conducted. Reflected power was measured using a logarithmic amplifier-based RF power detector (AD8307, Analog Devices, USA). The amplifier uses a progressive compression technique with 6 amplification stages. A narrow-band input matching at the frequency of interest also helps in better signal sensitivity along with a certain amount of frequency selectivity. As the output of the coupler is unbalanced, using unequal capacitor values in the input matching network provides a balanced differential drive at both the input ports of AD8307. The OFFSET feature allowed us to change the intercept point and further extend the dynamic range of the power detector. The lower range is largely limited by the thermal noise floor. Taking these considerations into account, the external matching circuitry was designed to provide the necessary dynamic range of 90dB at 300 MHz. The lumped element values C1 – 4 pF, C2 – 3.7 pF, and LM – 120 nH were calculated using ADS simulation using the touchstone (.s2p) file of the device from the manufacturer. The power detector has a slope of 25 mV per dB. A minimum analog output (DC value) of the detector was recorded to be 437 mV with a power supply and no RF signal. The analog output was fed to a 12-bit Analog-to-Digital (ADC) converter which was integrated in the microcontroller.
MEMS based matching network. Space inside a preclinical scanner is very limited and it was difficult to place the SWIM system inside the bore close to the RF coil without using a long coaxial cable. Therefore, we used a remote matching technique with two matching network boards [39–41]. An L-matching network with two trimmer capacitors (SG9128, Sprague Goodman, USA) close to the coil followed by a 2 m long coax cable extending out of the bore. We then connected the SWIM system outside the bore which includes a secondary L-matching network with two Single Pole 4 Throw (ADGM1304, SP4T, Analog Devices, USA) MEMS RF switches to create a capacitor array bank as shown in Fig. 3.
This switch was chosen because of high linearity and low COFF and RON for each port. The insertion loss and isolation of any one of the 4 ports (RFx) to the common port (RFC) at 300 MHz were measured to be 0.16 dB and 47 dB respectively. The switch also measured a crosstalk isolation of 45 dB between two ports. In shunt configuration, RFC was connected to the trimmer capacitor and each of the 4 switches were connected with a fixed capacitor (0505C Non-magnetic capacitors, Passive Plus Inc, USA) to ground. Similarly, in series arrangement, 4 fixed capacitors were connected to form a capacitor array. We also used a 10 MΩ resistor in shunt with each fixed capacitor to avoid floating capacitance during the off state of the switch. Each of the 4 throws can be individually controlled using Serial Peripheral Interface (SPI) in a multiplexer (MUX) style to create 256 combinations. The on-time of the MEMS switch was 75 µs, thus making the quickest time to cycle through all combinations without delays was approximately 20 ms. But a 5 ms delay between state transitions was introduced to account for RF settling time which makes a 3 second total run time for automatic tuning and matching sequence. The eight fixed capacitor values were picked by loading various sample sizes into the coil to understand the scope of impedance mismatch based on sample size. This method ensured that the SWIM system was not limited by capacitor values for a large variety of sample sizes.
Wireless controller. The wireless microcontroller (ESP32-WROOM-32D, Espressif Systems, China) used in the SWIM system is a powerful, cost-effective, and easy to use module with generic Wi-Fi, Bluetooth, and Bluetooth Low Energy (BLE) capabilities. We used the built-in 12-bit ADC of the ESP32 to read the reflected power from the power detector and map it to each of the 256 states of the matching network array. SPI protocol is used to control the MEMS switch instead of digital pins as serial communication is robust and makes it easier to customize the system by adding more MEMS switches in the future. An array of digital pins is connected to power detector and power supply regulators to enable and disable these devices according to need. This reduces additional power supply noise and reduces power consumption of the SWIM system during imaging sequence as these devices are turned off. An Android mobile application can be paired with the Bluetooth feature to create a user interface to interact with the microcontroller.
Android based application. We have developed an Android mobile application using MIT App Inventor as shown in Fig. 4a. After pairing the SWIM system to the mobile device, an ESP32 module can be connected individually with a unique MAC address. Up to eight individual SWIM systems can be connected to the Android application and controlled independently, thus allowing the user to scale this design to accommodate up to eight channel RF coils. Two buttons were designed to independently control the self-calibration module and AUTO T/M feature. Two input dialog boxes were placed to set the remote tuning and matching separately based on the coil parameters, known as the PRESET condition. Similarly, two slider bars were designed to provide wireless pseudo-manual control of tuning and matching capacitor array independently and irrespective of the auto-calibration function. After the automatic function, these slider bars can also be used to fine-tune the impedance condition. They also provide an easy wireless adjustment of impedance condition between scans to account for sample movement. Finally, we also included a ball animation, which stops moving after the automatic sequence completes, as a notification. Auto-calibration sequence is shown as a flowchart in Fig. 4b. Once a receive coil is connected to the SWIM system with a preset condition for the MEMS switches, an initial reading of reflected power is displayed for preset state and plotted on the black canvas as well. This is followed by loading of the sample and impedance matching of the transmit coil. To match the receive coil the main RF path was switched to self-calibration system and CW signal was enabled. When the AUTO T/M button is pressed, system cycles through all the possible combinations and plots the reflected power at the end of the automatic function. At the end, the system locks on to the least reflected power state. We confirmed the optimal state by manually changing states and checking the reflected power. Next, the self-calibration system is turned off and the main RF path is toggled back to the console receive chain, ready for imaging. After being satisfied with the impedance of the coil, Bluetooth was disconnected, and the microcontroller was placed in deep sleep to avoid adding any electronic noise to the image.