Introduction

One of the ultimate goals of fluorine-19 (¹⁹F) MRI of perfluorocarbon emulsions (PFCs) is the direct quantification and monitoring of inflammation in diseases such as atherosclerosis in patients in a clinical setting.^1,2 To this end, two different aspects of the atherosclerosis inflammation process were investigated, namely quantification and characterization of the inflammation in the plaques. The first aspect was an optimization of the inflammation quantification through wavelet denoising. Given that ¹⁹F is not naturally abundant in the body and that the concentration of injected PFCs is relatively low, especially in atherosclerotic plaques in mice where the signal is often buried in the noise, ¹⁹F MR imaging with a sufficiently high signal-to-noise (SNR) ratio is challenging. Therefore, besides optimizing the pulse sequence, a potential image quality improvement is to apply a wavelet denoising filter during the image reconstruction. Here, the image is transformed to the wavelet domain, where undesired noise is stored in low-magnitude coefficients, while signal is concentrated in large-magnitude coefficients. After shrinking the low-magnitude coefficients, the undesired noise is removed without affecting the signal.³ The second aspect was the characterization of the cell populations that take up PFCs in the inflammation process of the atherosclerotic plaques through imaging flow cytometry, in order to better understand the mechanisms of signal generation in ¹⁹F MR images. Therefore, we here aimed to: 1) demonstrate the feasibility of optimized ¹⁹F MRI in small atherosclerotic plaques at 3T, including quantification after wavelet denoising, and 2) characterize and visualize the PFC-incorporating immune cell populations involved in the inflammation process.

Methods

Images were acquired on a 3T clinical scanner (Prisma, Siemens) with a volume RF coil tunable to both ¹H and ¹⁹F frequencies (Rapid Biomedical). An ECG-triggered 2D ¹H GRE sequence was acquired for the localization of the plaques in the aorta and its branches, and an optimized 3D TSE pulse sequence⁴ (voxel size 0.78×0.78×1 mm³, echo train length 13, and TR/TE=1070/13 ms) was used for the acquisition of ¹⁹F images. C57BL/6 apolipoprotein-E-knockout (ApoE−/−) mice were fed with a high-fat diet to exacerbate atherosclerosis. Thirteen animals received 2×200μL of perfluoropolyether (PFPE, Celsense) intravenously, 1 and 2 days before imaging. An external reference with a fixed PFC concentration (CPFC=22mM) was placed next to the mouse and used for absolute quantification. Images were reconstructed with and without a wavelet denoising filter.⁵ The atherosclerotic plaques were segmented after ¹⁹F-¹H co-registration and thresholding. The signal intensity of each plaque (SI_plaque) relatively to the external reference was used in order to calculate the MRI-derived PFC concentration with both reconstruction schemes. The standard deviation of each SI_plaque (SD_plaque) was interpreted as an indicator of the precision in concentration measurement. The volume of the plaques (signal-volume) and the SI-signal-volume product (i.e. the ¹⁹F signal-integral) were also quantified. Five additional ApoE−/− mice were injected with fluorescein-isothiocyanate (FITC) PFPE and five control ApoE−/− mice were injected with non-fluorescent-PFPE for post-mortem imaging flow cytometry⁶ (ImageStream) to visualize all cells of the small aortic leukocyte populations.

Results

In all mice, ¹⁹F MR hotspots were detected in the aorta or the brachiocephalic artery (Figure 1). SNR in the plaques in images with the standard reconstruction scheme was 7.8±1.8. The MRI-derived PFC concentrations were 0.56±0.09mM and 0.49±0.10mM (P=0.001) for the standard and denoised images, while the plaque signal-volume varied from 4.5±1.0mm³ to 2.8±1.1mm³ for standard and denoised images (P<0.001), respectively. SD_plaque on average decreased by 47% from plaques of non-denoised ¹⁹F images to plaques of denoised ¹⁹F images. Among leukocytes, the PFC-labeled cells were mainly dendritic cells, macrophages and neutrophils at a ratio of 9:1:1. Small PFC spots were observed in macrophages (area=4.6±4.2µm²) and neutrophils (area=3.7±1.8µm²), as opposed to the dendritic cells (area=22.0±18.0µm², Figure 2). The PFC signal-integral (area×relative brightness) for dendritic cells was ~20× higher than that of macrophages and neutrophils.

Discussion and Conclusions

We demonstrated the feasibility of using ¹⁹F MR for the noninvasive quantification of PFCs at clinical magnetic field strength and in inflammation sites such as atherosclerotic plaques in mice. The lower SD_plaque confirmed that wavelet denoising allows for increased precision in concentration measurement in ¹⁹F images. The significantly lower plaque size and concentration obtained from the denoised images were most likely caused by the discarding of very small signals by the wavelet filter. Through imaging flow cytometry, we characterized the plaques immune cell populations and demonstrated that the advanced plaques studied here contained more dendritic cells than macrophages, which agrees well with previous studies.⁷ In conclusion, we have developed a quantitative ¹⁹F MRI technique for inflammation studies of low-SNR signals on a human MRI system that contributes to the translation of ¹⁹F MRI to the human setting.

Acknowledgements

This work was supported by grants from the Swiss National Science Foundation (PZ00P3-154719) to RBvH, as well as by the Centre d’Imagerie BioMédicale (CIBM) of the UNIL, UNIGE, HUG, CHUV, EPFL, and the Leenaards and Jeantet Foundations.