Sensitive detection of extremely small iron oxide nanoparticles in living mice using MP2RAGE with advanced image co-registration

Abstract

Magnetic resonance imaging (MRI) is a widely used non-invasive methodology for both preclinical and clinical studies. However, MRI lacks molecular specificity. Molecular contrast agents for MRI would be highly beneficial for detecting specific pathological lesions and quantitatively evaluating therapeutic efficacy in vivo. In this study, an optimized Magnetization Prepared-RApid Gradient Echo (MP-RAGE) with 2 inversion times called MP2RAGE combined with advanced image co-registration is presented as an effective non-invasive methodology to quantitatively detect T1 MR contrast agents. The optimized MP2RAGE produced high quality in vivo mouse brain T1 (or R1 = 1/T1) map with high spatial resolution, 160 × 160 × 160 µm³ voxel at 9.4 T. Test-retest signal to noise was > 20 for most voxels. Extremely small iron oxide nanoparticles (ESIONPs) having 3 nm core size and 11 nm hydrodynamic radius after polyethylene glycol (PEG) coating were intracranially injected into mouse brain and detected as a proof-of-concept. Two independent MP2RAGE MR scans were performed pre- and post-injection of ESIONPs followed by advanced image co-registration. The comparison of two T1 (or R1) maps after image co-registration provided precise and quantitative assessment of the effects of the injected ESIONPs at each voxel. The proposed MR protocol has potential for future use in the detection of T1 molecular contrast agents.

Figure 1

Schematic of experimental procedures. (a) Optimization of MP2RAGE for T1 mapping of in vivo mouse brain at 9.4 T (see Supplemental Figs. S1–S6 and Fig. 2). (b) Optimization of image co-registration (see Figs. 3, 4, 5). (c) In vivo detection of ESIONPs in R1 (= 1/T1) map (see Figs. 6, 7, 8).

Figure 2

MP2RAGE-based T1 MRI maps from 5 control mice at 160 µm isotropic spatial resolution in vivo. (a–e) Each column represents an individual mouse. Row i. Anterior coronal slices at the level of the genu of the corpus callosum. Row ii. Anterior coronal slices at the level of the hippocampal commissure. Row iii. Coronal slices at the level of the anterior hippocampus. Row iv. Coronal slices at the level of the posterior hippocampus. Row v. Posterior coronal slices at the level of the 4th ventricle.

Figure 3

Comparison of threshold-based vs. saturation-based co-registration. column i. 1st T1 map from the first scan. column ii. 2nd T1 map from the second scan. column iii. 2nd T1 map co-registered to the 1st T1 map. column iv. subtraction result of 1st T1 map from 2nd T1 map co-registered to the 1st T1 map. column v. difference between 1st T1 map and 2nd T1 map co-registered to the 1st T1 map. row (a) and (c) MR results from threshold approach with extreme T1 values set to zero. row (b) and (d) MR results from saturation approach with extreme T1 values set to 300 and 2400 ms (300 ms as shortest limit and 2400 ms as longest limit). The saturation based co-registration approach produced better image co-registration.

Figure 4

Absolute ΔT1 between first scan and second scan in 5 control mice. (a–e) Each column represents an individual mouse. Rows i, iii: absolute ΔT1 after co-registration using the threshold method. Rows ii, iv: absolute ΔT1 after co-registration using the saturation method. In all 5 mice, the saturation-based approach produced more accurate co-registration result than the threshold-based method.

Figure 5

Reproducibility of R1 (1/T1) mapping. After saturation-based co-registration, all T1 (s) maps were converted into R1 (s⁻¹) maps where R1 is defined as 1/T1. (a-i) R1 map using MP2RAGE. (a-ii) Repeat R1 map from a second acquisition in the same mouse, co-registered to the map from the first scan. (a-iii) ΔR1 map for each voxel, defined as R1 from scan 2–R1 from scan 1 after co-registration. (a-iv) test–retest signal to noise ratio (TrTSNR) map, defined at each voxel as the average R1 between the 2 scans divided by ΔR1. (b–e) TrTSNR maps for 4 additional mice. TrTSNR = 20 represents 95% reproducibility or 5% error. In all five control mice, TrTSNR of 20 or higher was observed.

Figure 6

Detection of Extremely Small Iron Oxide Nanoparticles (ESIONPs) stereotaxically Injected into the Mouse Brain in vivo by MP2RAGE MRI. All maps were zero-filled from 160 × 160 × 160 µm³ to 80 × 80 × 80 µm³. (a-i,b-i,c-i) R1 maps from pre-injection MP2RAGE scans of 3 individual mice. (a-ii) R1 map from MP2RAGE scan after injection of saline. (b-ii) R1 map from MP2RAGE scan after injection of 1 µl of ESIONPs at 0.1 mM iron concentration (0.1 nMol iron). (c-ii) R1 map from MP2RAGE scan after injection of 1 µl of ESIONPs at 0.25 mM iron concentration (0.25 nMol iron). (a-iii,b-iii,c-iii) R1 maps after injection co-registered to pre-injection. (a-iv,b-iv,c-iv) ΔR1 maps of pre-injection scan from post-injection scans co-registered to the pre-injection. (a-v,b-v,c-v) Absolute ΔR1 maps between pre-injection scan and post-injection scans co-registered to the pre-injection. The R1 enhancing effect of injected ESIONP is evident in both ΔR1 and absolute ΔR1 (|ΔR1|) maps.

Figure 7

Consistent detection of ESIONPs after injection. All maps were zero-filled from 160 × 160 × 160 µm³ to 80 × 80 × 80 µm³. Each panel represents for individual mouse, total 15 mice. (a–e) |ΔR1| maps after injection of 1 µl of saline in 5 individual mice. (f–j) |ΔR1| maps after injection of 1 µl of ESIONPs at 0.1 mM iron concentration in 5 additional mice. (k–o) |ΔR1| maps after injection of 1 µl of ESIONPs at 0.25 mM iron concentration in 5 additional mice. The box in panel (k) indicates the region of interest used for quantitative analysis. Note that small amounts of signal change ipsilateral to injection indicated by arrow can be detected, likely due to variability in injection technique.

Figure 8

Quantitative Analysis of ESIONPs detectability. |ΔR1| in a 6 by 6 voxel region of interest around the injection site (shown in Fig. 7k) plotted as a function of injected iron concentration. Dashed line shows the 95% confidence interval around the |ΔR1| for saline injection (0.0 mM [Fe]), and indicates the theoretical detection limit. Solid line indicates the linear regression between |ΔR1| and [Fe]. Linear regression intersects the theoretical detection limit at about 0.03 mM [Fe].