In the previous article in this series (Magnetic particle imaging: moving ahead), we took a look at work by researchers at the University of Lübeck in Germany in developing a single-sided MPI scanner and optimizing field-free line (FFL) scanning. The article also examined the development of new nanoparticles by a team at the University of Washington (Seattle, WA).

Alongside such advances in instrumentation and nanoparticle design, the final piece in the MPI puzzle could be considered as the development of improved image reconstruction techniques and hardware. Patrick Goodwill, a research associate working with Steve Conolly at UC Berkeley's Imaging Systems Laboratory (Berkeley, CA), told SPIE delegates about a new MPI theory: x-space MPI.

Moving to x-space

MPI involves the injection of superparamagnetic nanoparticles into the body. Unlike the iodine and gadolinium tracers used in X-ray and MR imaging, the nanoparticles used in MPI pose no risk to kidneys. The particles are subject to a strong static magnetic gradient, which saturates their magnetization everywhere besides a single field-free point (FFP). Drive coils then superimpose an oscillating magnetic field, and the nanoparticles' magnetic response is detected by a series of receive coils.

Only particles whose magnetic response is not saturated, i.e. those in the direct vicinity of the FFP, will contribute to the measured signal. Scanning the FFP through the sample enables reconstruction of a full tomographic image of particle distribution. Superparamagnetic nanoparticles exhibit a nonlinear response to an oscillating field, and standard MPI scanning uses higher-order harmonics in the detected signal to create images.

Currently, creation of an MPI image requires initial calculation of a system function matrix, which provides the relation between particle position and frequency response for a particular nanoparticle sample. Reconstruction is achieved through matrix inversion – a time-consuming processes that can lead to noise gain in the final image.

In contrast, x-space MPI involves recording an intrinsic MPI image, without any need for pre-characterization. X-space theory is based on three premises: that the magnetic gradient creates a single FFP; that particles exhibits a nonlinear magnetic response; and that the low-frequency information that's usually filtered out (as it is contaminated by the excitation signal) is recoverable.

During MPI scanning, as the FFP passes over a magnetic nanoparticle, the particle flips and induces a signal (in the form of a point spread function) in the receive coil. As the location of the FFP is known, a simple gridding process – essentially sampling the received signal onto a spatial grid that corresponds to the instantaneous FFP location – can transform the signal from the time domain to the image domain.

The final stage involves recovery of the low-frequency information using a partial FOV technique. The team demonstrated that overlapping partial FOV scans can be reconstructed into a smooth and contiguous image over the entire FOV, and that this process enables recovery of the lost fundamental signal without adding significant noise.

"X-space theory accurately describes how to acquire an intrinsic MPI image," explained Goodwill. "This means that there's no requirement for harmonics, system functions or matrix inversion."

Initial images and hardware

Goodwill and colleagues tested their x-space theory by constructing two scanners: a high-resolution small-scale MPI scanner, and the first FFL MPI scanner, also presented at the SPIE conference.

The high-resolution scanner (8.9 cm magnet free bore, 6 T/m gradient, 4 x 2 cm FOV) was used to image a "CAL"-shaped phantom made from 400 µm tubing filled with Resovist. With an imaging time of 28 s, the system created images with a resolution of approximately 1.6 mm.

The researchers claim that these represent the first MPI images of a complex phantom without any sharpening or deconvolution and with full recovery of the fundamental frequency. Goodwill noted that while deconvolution is not necessary, it can be used to improve the image if required.

The FFL scanner is sized for imaging mice. The system has a 10 cm magnet free bore, a gradient strength of 2.35 T/m and a 5 x 12 cm FOV. They are now working on a higher-resolution version for imaging rats, with a 13 cm magnet free bore and a gradient strength of 12 T/m. This system is expected to achieve better than 1 mm resolution with Resovist.

Since the SPIE meeting, the team has used the FFL scanner to image a laser-cut acrylic phantom with a line width of approximately 1.2 mm. The resulting image (see photo) – taken with a 4.5 x 12 cm FOV and 18 s acquisition time – is said to be the world's first projection MPI image. "From what we see in the literature and what we saw at the SPIE conference, we believe that it is the finest MPI image ever taken," Goodwill told medicalphysicseweb.

• If you've not seen it already, don't forget to check out the first article in this two-part series: Magnetic particle imaging: moving ahead.