INTRODUCTION
One developing application of magnetic nanoparticles is to use magnetically linked drugs in intra-arterial chemotherapy (IAC) procedures (1). These magnetic drugs can then be selectively removed by deploying an endovascular magnetic device downstream of the targeted organ (1, 2), thus limiting off-target drug toxicities. In vitro studies used radiolabeled iron oxide nanoparticles (IONP) to quantify the number of particles captured on the device using a gamma counter (2). To demonstrate efficacy in vivo, accurate quantification of the drug’s distribution on the device, within the targeted organ, and systemically, is necessary. The purpose of this study is to validate PET/MRI image-based quantification, using 89Zr-PET signal, transverse relaxation rate (R2*), and quantitative magnetic susceptibility (χ) of ferromagnetic 89Zr-IONP biodistribution in vitro and in vivo.
METHODS
In vitro experiment
Iron oxide (Fe3O4) nanoparticles with 50–100 nm diameter were used (Sigma-Aldrich, St. Louis, MO). Zirconium-89 (1.01 mCi, 2 μL) in 1M oxalic acid (3D imaging, Little Rock, AK) was added to IONPs (16.75 mg, 200 μL) in deionized (DI) water and 2 μL of 1M Na2CO3. The mixture was heated at 120˚C for 2 hours, and the 89Zr-IONPs were purified using a magnet. A phantom was constructed using a cylinder of gelatin with five embedded plastic spheres with increasing 89Zr-IONP concentrations of [2.5, 5, 10, 20, 30] mg/L. We used a 3.0T PET/MR (SIGNA, GE Healthcare, Waukesha, WI), with a head and neck coil with 12 channels for simultaneous acquisition of a 10-minute time-of-flight (TOF) static PET and MR data. A 3D multi-echo gradient echo UTE research sequence with a cones k-space trajectory (3) was used with TE = [0.148, 3.3, 5.6, 7.9] ms, TR = 13.2 ms, flip angle 12°, matrix 128×128, slice thickness 3 mm, field of view (FOV) 16×16 cm2, a coronal scan plane, a fat-suppression pulse applied every five TRs, and slab-selection.
In vivo experiment
With Institutional Animal Care and Use Committee approval, a 5 French catheter (Slip-Cath® Beacon® Tip, Cook Medical, Bloomington, IN) was placed into the common hepatic artery of a supine swine (45.8 kg, female), using X-ray guidance (Cios Alpha, Siemens Healthineers, Munich, Germany). Radiolabeling was the same as before, using 3.05 mCi 89Zr per 25 mg IONPs in 10 ml saline, shaken and separated into three aliquots: 1, 4, and 5 ml. We imaged before infusion, moved the scanner table away from the magnet bore for infusion, then imaged after each infusion. We used the same sequences as the in vitro study, except with an axial scan plane, 10 mm slice thickness, TE = [0.148 1.8, 3.6, 5.4] ms, TR = 10.2 ms, FOV 26×26 cm2, and an 18-channel upper anterior array coil and posterior array to acquire a breath-held sequence.
Data analysis
PET images were reconstructed using a TOF ordered subsets expectation maximization algorithm in a 256×256 matrix, 5 iterations, 28 subsets, point-spread-function model, and a standard z-axis post filter with cutoff 3.0 mm full-width at half maximum. We used non-attenuation correction (NAC) images in vitro, and attenuation-corrected (AC) images in vivo. Regions of interest (ROI) were manually drawn on an MR image slice (OsiriX v.6.0.2, Pixmeo SARL, Geneva, Switzerland). Post-processing was performed in MATLAB 2016a (MathWorks Inc., Natick, MA). PET ROIs were interpolated to match the MR resolution for pixel-wise analysis. R2* was measured using a mono-exponential fit of magnitude images. Quantitative susceptibility maps (QSM) were reconstructed using the Morphology Enabled Dipole Inversion (MEDI) toolbox (4–7). In vitro, we used a spherical mean value operator of radius 5, a regularization parameter of λ = 101.25, a scaling factor of 2.5303 and a global shift of −0.2554 ppm. In vivo, the same values were used, except using no scaling, λ = 104 pre-infusion, and λ = 102.5 post-infusion. A linear regression model was performed for the mean of each parameter ROI and IONP concentration, and then pixel-wise between each contrast mechanism.
RESULTS
In vitro results
There was a significant linear relationship with good correlation between 89Zr-IONP concentration and the mean for each parameter (Fig. 1, Table 1). Each parameter also demonstrated a significant linear relationship with each other pixel-wise, which was well described by the least-squares fit line (Fig. 1c).
Figure 1. In vitro experiment.
The a) magnitude image (second TE = 3.3 ms) annotated with the IONP concentrations (mg/L), 89Zr-PET fused on the magnitude image, R2* map, and χ map are shown for a single middle slice. b) The mean and standard deviation of a manually drawn ROI, in each gel ball are plotted for each parameter, and the means were then fit with a least-squares fit line. c) Comparing the parameters with each other, a least squares line was fit between 89Zr-PET and R2*, χ and 89Zr-PET, and χ and R2*.
Table 1.
The mean (± standard deviation) of the iron oxide nanoparticle phantom concentrations for each parameter, 89Zr-PET, R2*, and χ and the linear regression results for the five mean datapoints versus concentration are reported.
| [IONP] (mg/L) | [IONP] (mM) | 89Zr-PET (a.u.) | R2* (s−1) | χ (ppm) |
|---|---|---|---|---|
|
| ||||
| 2.5 | 0.011 | 2304.3 ± 272.9 | 107.1 ± 18.4 | 0.61 ± 0.03 |
| 5 | 0.022 | 2453.4 ± 342.5 | 110.7 ± 14.5 | 0.58 ± 0.03 |
| 10 | 0.043 | 3591.1 ± 460.6 | 164.8 ± 34.7 | 0.81 ± 0.16 |
| 20 | 0.086 | 4778.9 ± 597.3 | 240.5 ± 47.2 | 1.01 ± 0.20 |
| 30 | 0.130 | 6376.4 ± 837.6 | 302.7 ± 46.2 | 1.13 ± 0.17 |
|
| ||||
| Slope (±SE) | 149.0 ± 7.5 | 7.4 ± 0.4 | 0.021 ± 0.003 | |
| Intercept (±SE) | 1888.7 ± 127.0 | 84.9 ± 6.8 | 0.548 ± 0.048 | |
| Adjusted R2 | 0.990 | 0.989 | 0.930 | |
| p-value | 0.0003 | 0.0004 | 0.005 | |
In vivo results
Mean 89Zr-PET, R2*, and χ showed good correlation with a linear relationship to concentration, but only R2* demonstrated significance (Fig. 2, Table 2). 89Zr-PET and R2* showed the best correlation with each other, and lower correlations were shown between χ and R2* and χ and 89Zr-PET (Fig. 2e).
Figure 2. In vivo experiment.
A single middle slice is shown prior to infusion and after each infusion for the a) magnitude images (third TE = 3.6 ms), b) 89Zr-PET fused on the same magnitude images, c) R2*, and d) χ maps. e) Comparing the parameters with each other, a least squares line was fit between 89Zr-PET and R2*, χ and 89Zr-PET, and χ and R2*.
Table 2.
The mean (± standard deviation) of the liver ROI before intra-hepatic infusion, and after each infusion, for each parameter, 89Zr-PET, R2*, and χ and the linear regression results for the five mean datapoints versus estimated concentration are reported.
| IONP (mg/L) | 89Zr-PET (Bq/ml) | R2* (s−1) | χ (ppm) |
|---|---|---|---|
|
| |||
| Pre-infusion: 0 | - | 67.6 ± 29.3 | −0.033 ± 0.155 |
| Post-infusion 1: 2.5 | 4,544 ± 2,827 | 82.7 ± 23.8 | −0.042 ± 0.143 |
| Post-infusion 2: 12.5 | 22,347 ± 13,341 | 112.9 ± 34.1 | −0.027 ± 0.219 |
| Post-infusion 3: 25.0 | 69,627 ± 42,924 | 216.2 ± 90.0 | 0.071 ± 0.528 |
|
| |||
| Slope (±SE) | 2929.1 ± 568.5 | 5.75 ± 0.89 | 0.0042 ± 0.001 |
| Intercept (±SE) | −6881.2 ± 9210.3 | 62.3 ± 12.5 | −0.050 ± 0.020 |
| Adjusted R2 | 0.927 | 0.931 | 0.740 |
| p-value | 0.122 | 0.023 | 0.090 |
DISCUSSION
All three parameters, 89Zr-PET, R2*, and χ, had a significant linear relationship with 89Zr-IONP concentration in vitro, which were well explained by the least-squares fit equations. 89Zr-PET and R2* had the highest correlations with 89Zr-IONP concentration, followed by χ. 89Zr-PET and R2* had the best correlation with each other, then χ and 89Zr-PET, and then χ and R2*. In vivo, there was a significant linear relationship between the 89Zr-IONP concentration and R2*, but not χ or 89Zr-PET. In vivo, 89Zr-PET and R2* had the best correlation with each other, whereas χ and R2*, and χ and 89Zr-PET had lower correlation coefficients.
While this paper includes validation in a phantom and feasibility in vivo, this study has limitations. The liver has unique challenges that affect QSM accuracy, including chemical shift caused by fat. A water-fat separation method would provide more accuracy (8). Fat-saturation appeared slightly inhomogeneous. It may also suppress some water signal, leading to lowered SNR and R2* estimates (9), so our R2* rates may be low. ΔR2* may be required for broader ranges. It was difficult to achieve a homogenous IONP distribution in the phantom and liver. The nanoparticles are used as a proxy for magnetically linked drugs; the drug delivery mechanism needs to be optimized for hepatic delivery, since malignant liver tumors do not uptake superparamagnetic iron oxides (10). Since the endovascular filtration device is magnetic, it cannot be used in the scanner, and PET/MRI quantification cannot be performed real-time during the infusion; co-registration between scans will be needed.
In conclusion, this study demonstrated feasibility of image-based quantification of radiolabeled magnetic nanoparticles using PET/MRI in vitro and in vivo for a multi-modality assessment of 89Zr-IONP biodistribution.
Acknowledgments
The authors gratefully acknowledge Dr. Hongjiang Wei and Dr. Chunlei Liu for insightful discussions about susceptibility mapping, Emma Bahroos for PET-MR expertise, Vahid Ravanfar for PET-MR scanning, Dr. Michael Carl for assistance with the UTE sequence, and Dr. Pauline W. Worters, Dr. Mehdi Khaligi, and Patrick Koon for assistance with GE pulse sequences.
Grant Support: This publication was supported by NIH Grant R01 CA194533, NIH Grant R01 EB012031, a UCSF Radiology Departmental Seed Grant, and the Prescient postdoctoral fellowship from the National Center for Advancing Translational Sciences, NIH, UCSF-CTSI Grant Number TL1 TR001871. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.
References
- 1.Mabray MC, Lillaney P, Sze C-H, et al. In Vitro Capture of Small Ferrous Particles with a Magnetic Filtration Device Designed for Intravascular Use with Intraarterial Chemotherapy: Proof-of-Concept Study. J Vasc Interv Radiol. 2016;27:426–432e1. doi: 10.1016/j.jvir.2015.09.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kondapavulur S, Cote AM, Neumann KD, et al. Optimization of an endovascular magnetic filter for maximized capture of magnetic nanoparticles. Biomed Microdevices. 2016;18:109. doi: 10.1007/s10544-016-0135-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Gurney PT, Hargreaves BA, Nishimura DG. Design and analysis of a practical 3D cones trajectory. Magn Reson Med. 2006;55:575–582. doi: 10.1002/mrm.20796. [DOI] [PubMed] [Google Scholar]
- 4.Liu T, Wisnieff C, Lou M, Chen W, Spincemaille P, Wang Y. Nonlinear formulation of the magnetic field to source relationship for robust quantitative susceptibility mapping. Magn Reson Med. 2013;69:467–476. doi: 10.1002/mrm.24272. [DOI] [PubMed] [Google Scholar]
- 5.Liu J, Liu T, De Rochefort L, et al. Morphology enabled dipole inversion for quantitative susceptibility mapping using structural consistency between the magnitude image and the susceptibility map. Neuroimage. 2012;59:2560–2568. doi: 10.1016/j.neuroimage.2011.08.082. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Liu T, Liu J, De Rochefort L, et al. Morphology enabled dipole inversion (MEDI) from a single-angle acquisition: Comparison with COSMOS in human brain imaging. Magn Reson Med. 2011;66:777–783. doi: 10.1002/mrm.22816. [DOI] [PubMed] [Google Scholar]
- 7.de Rochefort L, Liu T, Kressler B, et al. Quantitative susceptibility map reconstruction from MR phase data using bayesian regularization: validation and application to brain imaging. Magn Reson Med. 2010;63:194–206. doi: 10.1002/mrm.22187. [DOI] [PubMed] [Google Scholar]
- 8.Lin H, Wei H, He N, et al. Quantitative susceptibility mapping in combination with water-fat separation for simultaneous liver iron and fat fraction quantification. Eur Radiol. 2018 doi: 10.1007/s00330-017-5263-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Meloni A, Michael Tyszka J, Pepe A, Wood JC. Effect of inversion recovery fat suppression on hepatic R2* quantitation in transfusional siderosis. Am J Roentgenol. 2015;204:625–629. doi: 10.2214/AJR.14.12786. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Duda SH, Laniado M, Kopp AF, et al. Superparamagnetic iron oxide: detection of focal liver lesions at high-field-strength MR imaging. J Magn Reson Imaging. 1994;4:309–314. doi: 10.1002/jmri.1880040314. [DOI] [PubMed] [Google Scholar]