To investigate the use of 4D spectroscopic SWIFT [1] for invivo assessment and quantification of IONP bio-distribution, and to compare it with the previously reported longitudinal relaxation time constant (T₁) assessment in mice.

Iron oxide nanoparticles (IONPs) are receiving increased attention for their potential application in theranostics (act as both diagnostic and therapeutic agents) [2]. A noninvasive and quantitative imaging techniques for the assessment of IONP bio-distribution invivo is crucial for effective and safe treatment. To expand the detectable range to clinically relevant levels, which falls into the gap between conventional MRI and CT [3], we have suggested a positive-contrast based technique using SWIFT Look-Locker T₁ mapping method [3-5]. However, cell internalization of IONPs will usually influence the relaxivity r₁. It has been reported the influence on r₂ and susceptibility is much less significant [6]. Therefore, for more accurate assessment and quantification, in supplement to the previous reported T₁ technique, a 4D spectroscopic SWIFT technique [1] was applied to characterize IONPs in mice invivo up to high concentration.

Spectroscopic SWIFT is based on the concept of adding a pseudo gradient to x, y, z gradients, to create an intrinsic spectroscopic frequency distribution [7]. Ideally, to reconstruct a 4D image, we must collect a set of projections with orientations isotropically distributed on a “4D sphere”. We achieved this by using a set of shrinking spatial spheres accordingly with decreased acquisition bandwidth (Fig.1). The 4D dataset was then reconstructed using the gridding method [9] after correlation step for SWIFT sequence.

Invivo studies were done on 3 nude mice. The superparamagnetic IONPs (EMG-308, Ferrotec, USA) coated with mesoporous silica and polyethylene glycol at concentrations of 0.16, 0.17, 0.18 mg Fe/(g of body weight) were delivered by intravenous (IV) injection. (1 mg Fe/ml = 17.8 mM Fe). Normal MB-SWIFT [8] images and 4D spectroscopic SWIFT images were acquired 1 week after IV injection on a 9.4 T animal MRI scanner (Agilent Technologies, USA) using a volume coil. MB-SWIFT: BW=384 kHz. Spectroscopic SWIFT: BW_max=125 kHz, bandwidth in frequency dimension Ω =15 kHz, total number of projections = 41792, acquisition time = 12 minutes. GRE images were also acquired with BW=150 kHz, TR=4.2 ms, TE=2.1 ms. The spectrum of each voxel was fitted to Lorentzian function to get T₂*. [Fe] in organs were measured by ICP-MS.

Images of post-injection mouse from the same slice, but at different frequencies are shown in Fig. 2, along with GRE and MB-SWIFT images. After injection, most IONPs deposited in liver, spleen, and kidney. These organs appeared as a void in GRE images due to the ultra-short T₂* caused by IONPs at high concentration, while there are still plenty of signal in MB-SWIFT images. The presence of IONPs changes the water resonance frequencies of surrounding tissues. These were observed in spectroscopic SWIFT images at different frequencies (orange arrows fig.2) and in the spectrum over marked profile (red arrows fig.2). The B0 and susceptibility information can be further extracted. The image at fat frequency illustrated the peri-renal fat clearly (yellow arrows Fig. 2).

The spectroscopic images of a mouse body before and after IONPs-injection showed in left column of Fig.3. The spectroscopic information of liver, spleen and kidney, and the fitted T₂* down to 20 microseconds were obtained (right part of Fig.3). Apparent line broadening effect was observed in the post-injection case for all three organs. In spleen, a binary distribution giving two very different T₂* values was observed (Fig.3f), which was verified by histology showing majority of IONPs packed in red pulp rather than white pulp (Fig.4a and 4b). This can also be seen at MB-SWIFT images (Fig.4c). The acquired T₂* map is shown in Fig.4d. Notice the fitted T₂* values in spleen regions (white arrow heads) reflect the averaged values from both red and white pulps. A bi-component fitting could be used to fit components separately. Due to the limited resolution in the frequency dimension (Δf = 230 Hz), tissues with T₂* higher than 1.5 ms cannot be accurately estimated. Compressed sensing will be used to increase frequency resolution while keep acquisition time in a reasonable range in future.

A 4D spectroscopic SWIFT technique was applied to characterize the bio-distribution of IONPs in vivo in mouse. Resonance frequency shift from susceptibility difference and T₂* shortening effect (down to 20 microseconds) caused by IONPs was observed in liver, spleen and kidney. The acquired T₂* map provides quantitative bio-distribution information of IONPs making the 4D spectroscopic SWIFT a promising tool in nanoparticle-based theranostics.