MPI_Tx_Rx_Coils
The combination of a transmit (also referred to as "drive" and "Tx") and receive ("Rx") coils form the basis of signal production and acquisition in MPI. Drive coils excite the SPIONs within the FOV with high amplitude AC magnetic fields (~12mT at 25kHz) and the receive coil acquires the signal while attenuating signal being generated directly from the drive coils.
Figure 1: A 3/4 rendering of the Tx/Rx assembly showing the coil arrangement and structural support.
For the following coils, we have an overall free bore diameter of 50mm. This is to ensure a rat head can fit into our scanner.
To excite the particles within the FOV, there is a singular goal which is to create a drive field at sufficient amplitude. The word "sufficient" is intentionally left vague as the drive amplitude is heavily informed by the application. For instance, if you planning on modifying the small-bore imager for minimum power such as Graeser et al. does for their system, a drive field of 6-8mT might be "sufficient"(1), but we use 12mT. The 12mT value comes from our particles (Vivotrax, Magnetic Insight, Alameda, CA) having their saturation value (and thus the point where substantial non-linearities come about) approximately 12mT. By driving higher amplitudes you gain SNR, but interestingly has been shown to lower spatial resolution(2). Further, if you build the MPS system, it is feasible to characterize the particles you plan to use and then optimize the excitation field amplitude based on those data.
The everpresent design limitation is heating. Heat is the name of the game with power electromagnet design. With high amplitudes, you get substantial thermal drift (especially without water cooling) and potentially coil damage if the heat really gets out of hand. Our current drive coil is passively air-cooled and seems to work well, but the thermal drift is definitely substantial, so adding water cooling is on the list of future improvements. Malhotra et al. describe an interesting method for water cooling in their 2019 INSPECT paper(3). Keep in mind that other resistive elements in the drive current path with heat too. For our toroidal inductors in the Tx filter, we water cool them by wrapping water tubes around the inductors and flowing cold water. This approach is simple to implement and has almost zero leaking risks.
The most significant influence on the heating optimization question is wire diameter-- thin wire allows for higher turn density which means lower requisite current for a given magnetic field. But for a given current, heat density is proportional to diameter to the fourth power (Resistance is proportional to diameter squared, power is proportional to resistance squared), yet there is a geometric aspect to design to consider as well such as proximity effect and ensuring the coolant can sufficiently access the coils.
With that said we have designed the following drive coil:
The purpose of the filter is to minimize the impedance of the drive frequency and simultaneously maximize the harmonic attenuation. Therefore, we currently use a tuned low-pass filter which has a resonance at 25kHz and 84dB attenuation by the third harmonic at 75kHz. The filter is enclosed within a shielding box constructed out of aluminum. The shielding is mainly to keep the high-power 25kHz signal from radiating into the nearby receive chain elements. The thickness of this box is 12AWG (~2.0mm) and with the skin depth at 25kHz in aluminum being ~0.5mm, this results in over 95% attenuation of fields from radiating through the box due to the eddy current shielding effect.
The Rx coil in some ways has more subtle design decisions involved. Below is our design but keep in mind there are certainly options to explore. For one, wire diameter is a pretty major question for different (yet similar) reasons. There is the tradeoff of turns versus resistance again but this time the resistance plays a role because eventually, thermal noise can limit the sensitivity. One option we have discussed (and are testing on the MPS platform) is designing the coil so the coil thermal noise is equal to the pre-amp noise.
The Rx filter is a component that is currently under debate. Either it can be designed to resonate at F3 to improve the sensitivity of the system, although this approach has a risk for reducing spatially relevant information. The other approach more often used within the field of MPI is to use a broadband receive chain and a notch filter to reduce drive feedthrough. A third option is to receive broadband without notching out the drive frequency, and attenuating it purely by geometric decoupling ('Geometric decoupling' refers to the gradiometer canceling out the homogeneous field at the drive frequency.) with active cancellation further attenuating drive coil feed-through. We currently use a filter which is resonant at the third harmonic.
We use commercial pre-amplifiers coupled together for low-noise signal amplification and anti-alias filtering. Specifically, we have an Ametek 5113 that goes into a Stanford Research System SR560. These systems have quoted noise generation on the order of ~4nV/sqrt(Hz).
(1) Graeser et al. Human-sized magnetic particle imaging for brain application, 2019
(2) Croft et al. Low drive field amplitude for improved image resolution in magnetic particle imaging, 2015
(3) Malhotra et al. Tracking the Growth of Superparamagnetic Nanoparticles with an In-Situ Magnetic Particle Spectrometer (INSPECT)