Home
Figure 1 :A digital rendering of the small-bore imager done in Inventor
The imager employs the FFL-based MPI imaging scheme as FFL-MPI has higher sensitivity when compared to FFP MPI due to receiving signal from a greater region at any given instant (1). We transmit through a solenoidal drive coil through a tuned low pass filter. We receive through a solenoidal pair gradiometer coil. Currently, the signal is passed through a filter that notches out Tx feedthrough and resonates at the third harmonic.
Due to using permanent magnets to generate the FFL, we must use a physically rotating gantry to enable taking projections around the bore. Further, as we are largely interested in looking at functional activity with temporal changes every ~5 seconds, we chose to have this gantry be continuously rotating as opposed to stopping and reversing after each image. Facilitating rotation, there is an electrical slip ring to transmit electrical power to the shift coils and a water rotary union (RU) to transmit coolant to the shift coils. The other main mechanical consideration is the sample bed. The bed is controlled via a gear-reduced stepper motor coupled to the bed via a timing belt.
Gradient: 2.83 T/m
Drive field: 12mT
FOV: ~30mm diameter
Resolution: ~2mm
Temporal Resolution: Under 3 seconds
Sensitivity: 50 ng in 75 seconds
We employ an array of four N48 NdFeB permanent magnets to form each one of the two opposing FFL magnets. The dimension of each array is 2” x 2” x 16” consisting of four 2” x 2” x 4” blocks that have been glued together and are contained within an ABS enclosure. The magnets are located ~5" apart (when measured between their inner faces), this separation distance inherently determines the gradient strength, shift coil requirements, bore diameter, etc. and due to the wide-reaching influence of this design parameter, it has its own page here.
Note: the coils described are currently being constructed. The data previously collected is with a different shift coil geometry described in the shift coil's main page We have chosen to use “racetrack” shaped shift coils assembled in four-layer sub-assemblies separated by heat sinks. Each layer is composed of 20 turns of 10-gauge solid magnet wire. The geometry was determined by the requirement to make the coils go around the FFL gradient magnet, and designing for a non-square cross-section because flatter geometries enable more efficient heat removal due to greater surface area. Current in the coils was determined by knowing the FFL must be shifted the entire FOV (up to 15 mm each way). With a gradient of 2.83T/m and a field efficiency of ~1mT/A, to shift 15mm you need 42.5 Amps (peak).
The drive coil is a two-layer, passively cooled, solenoid coil constructed of 16AWG Litz wire. The drive coil is wholly encased in epoxy to stabilize the wires during operation. We designed the drive coil to reach ~12mT within the core.
We use a tuned low-pass filter which has a resonance at 25kHz and 84dB attenuation by the third harmonic at 75kHz (from simulation). All of the toroidal inductors are water-cooled and the capacitors are Celem power capacitors. All steel components have been removed from the Tx chain including threaded inserts which originally came in the power capacitors to mitigate the risk of non-linearities being introduced within the high-power transmission. The filter is enclosed within a shielding box constructed out of aluminum.
The Rx coil is designed as a pair of oppositely wound coils to minimize the amount of the drive field inducing a net voltage in the Rx chain. This coil is wound from Litz wire to reduce the thermal noise generated from the coil.
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. Yet even this is not necessarily the “correct” approach as it has been shown to be possible to attenuate feed-through via passive and active means and use the true signal at the drive frequency to boost the SNR as D. Pantke et al. suggest in their recent paper.
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).
The “rotating gantry” encompasses all of the rotating components of the system as well as the slip ring/rotary union. For the most part, the system is constructed of fiberglass sheets (FR4) with 3D printed elements to help support it and a large timing belt pulley which serves as the method for turning it. The structure supports the FFL magnets, the shift coils, and the water cooling with the primary mechanical loading being due to the weight forces, magnetic repulsion, and the torque required to overcome the friction inherent in the rotary union, where the rotary union’s torque, being dynamic friction, requires substantial energy from the motor to overcome. Regarding material selection throughout, polymer-based materials are preferred because conductive components will generate large eddy currents from the shift coils, and although these eddy currents don’t make the system dysfunctional, it will cause an increased phase-shift in the shift field (with respect to applied voltage) and diminish the efficiency. To overcome this, when we used metallic components such as in heat sinks, we added a dielectric break if possible.
The stationary frame is constructed from extruded aluminum framing and besides supporting the gantry and the stationary tube should play a role in vibration management.
The internal coils consist of the Tx/Drive coil and the Rx coils/gradiometer which are epoxied together. The coil assembly is then set so the gradiometer coil nearest to the forward end of the coil (where the rodent/sample would enter) is located at isocenter. They are held in place by a series of wedges that engage via threaded rods. The other key component within the copper tube is the bed/rail assembly. The rails enable consistent, stable travel down the bore. The travel is driven by a stepper motor which is coupled to the bed through a timing belt. The “bed” can either be a platform designed for an anesthetized rat, or SPION filled phantoms.
(1) Weizenecker et al. Magnetic particle imaging using a field free line, 2008 [Link](https://iopscience.iop.org/article/10.1088/0022-3727/41/10/105009/meta)