1.

Introduction

During the past decade, the concept of x-ray luminescence imaging (XLI) emerged and demonstrated great potential for molecular imaging in small animals by combining the high-spatial resolution of traditional x-ray imaging and the superb measurement sensitivity of optical imaging. Specifically, x-ray luminescence computed tomography (XLCT) is a powerful imaging modality capable of the high-resolution imaging of deeply embedded x-ray excitable contrast agents in 3D. In principle, a focused or collimated beam of x-ray photons are used to penetrate deeply through the tissue samples, with minimal scatter; x-ray excitable contrast agents within the x-ray beam path absorb the x-ray energy, causing a photophysical cascade of events, eventually leading to the emission of many optical photons, which can then pass through the tissue and escape from the skin. Highly sensitive photodetectors, such as electron multiplying charge-coupled device (EMCCD) cameras or photomultiplier tubes (PMTs), can then detect the optical photons for image reconstruction. Pratx et al.^1,2 first reported a selective-excitation-based XLCT imaging and demonstrated that the distribution of deeply embedded contrast agents could be recovered with high-resolution and sensitivity with this method. Soon following, several research groups, including our own, improved several aspects of XLCT including the XLCT imaging systems,^3–13 robust reconstruction algorithms,^4,14 and designing efficient, bright, and biocompatible XLCT imaging probes.^15–20

XLCT can be performed using different x-ray beam geometries, each with their own advantages and disadvantages. Selective beam-based XLCT is performed using a narrow x-ray beam that raster scans across the sample such as the first-generation computed tomography (CT) scanners. This allows for very high-spatial resolution since the beam size and location can be used as a priori information in the image reconstruction. The drawback of this method is the inherently long scan time due to the small excitation region and time needed to traverse the whole object. Another geometry is the fan or planar beam geometry, which can be used to image an entire cross-section at one time and allows for selective-planar imaging.^11,12 This method eliminates the need for raster scanning, which improves the imaging time; however, the lateral spatial resolution is degraded compared with the narrow beam selective excitation since the excitation is no longer localized to a small region but rather the entire cross-section. The last x-ray beam geometry is the conical-beam geometry, which offers the fastest scanning time of the three geometries mentioned since the object is entirely excited by the beam at one time.¹³ However, due to the ill-posedness of the image reconstruction, this method further compromises the spatial resolution as a trade-off. Due to advantages of the improved spatial resolution, our efforts have focused on the narrow x-ray beam-based XLCT imaging.

We performed narrow x-ray beam-based XLCT imaging using collimators to generate a fine x-ray beam and used an EMCCD camera to detect the optical photons that reach the object surface.^3–6,21 With this method, we achieved promising results and demonstrated that we could separate targets with an edge-to-edge distance of 0.15 mm²² and could perform XLCT imaging at depths greater than 2 cm and at concentrations of $0.01\mathrm{mg}/\mathrm{ml}$ (${\mathrm{Gd}}_2{\mathrm{O}}_2\mathrm{S}:{\mathrm{Eu}}^{3+}$ contrast agent).²¹ To improve on reducing the imaging time, a challenge for the narrow-beam-based XLCT, we also proposed the idea of using a multiple-pinhole collimator to generate multiple scanning x-ray beams and thereby reduced the imaging time by a factor equal to the number of beams.⁶ More recently, we have explored using both a higher-flux x-ray beam and more sensitive optical detectors to perform XLCT imaging. Collimation, while easy to implement, is highly inefficient as most photons generated by the x-ray source are absorbed by the collimator, with only a small number of photons able to pass through the pinhole. We recently demonstrated that we can increase the x-ray photon flux, and thus perform much more sensitive XLCT using x-ray optics (e.g., a polycapillary lens), to capture divergent x-ray photons from the source and refocus them with a dual-cone geometry. Also, using PMTs instead of an EMCCD camera, we can acquire our optical signal more rapidly and with a higher signal-to-background ratio.⁷

XLCT relies on x-ray luminescence scintillators to generate light in the tissue with minimal background from the endogenous tissue. Rare earth-doped nanophosphors are promising XLCT contrast agents for noninvasive bioimaging, such as to detect tumors in vivo. For such applications, the nanophosphors should ideally be small, colloidally stable, and chemically functionalizable for targeting, and generate bright radioluminescence signals with tissue penetrating red or near-infrared (NIR) light. Many microphosphors are commercially available, mostly developed for lighting, display, and x-ray imaging applications. Smaller stable nanophosphors have also been developed for biolabeling applications, especially fluorescence and upconversion, and some of these (e.g., ${\mathrm{NaGdF}}_4:\mathrm{Eu}$ and ${\mathrm{Gd}}_2{\mathrm{O}}_2\mathrm{S}:\mathrm{Eu}$) are also good XLI contrast agents.^2,15–20 Reducing the phosphor size usually decreases their luminescence yield due to quenching from surface defects and escape of x-ray generated photoelectrons, but annealing processes can help by reducing quenching from defects. Based on our previous work with XLCT imaging, along with the efforts of our collaborators, we have proposed and are currently building a small-animal dedicated focused x-ray luminescence tomography (FXLT) imaging system that also incorporates a micro-computed tomography ($\mu \mathrm{CT}$) scanner. We also synthesized nanophosphors with various emission wavelengths to be used as contrast agents for multiplexed XLCT imaging. Lastly, we propose a deep-learning-based XLCT reconstruction algorithm to improve upon our existing reconstruction methods. The remainder of the paper is arranged as follows. In Sec. 2, we discuss the design and build of our designed FXLT imaging system and the proposed scanning scheme. In Sec. 3, we discuss the proposed synthesized nanophosphors. In Sec. 4, we discuss the deep-learning-based XLCT reconstruction algorithm and numerical simulations. Lastly, Sec. 5 presents the summary and conclusion.

2.

FXLT Imaging System Design and Scanning Scheme

The proposed FXLT scanner was designed by computer-aided design (CAD) software and built at our Biomedical Imaging Lab at the University of California, Merced, as shown in Figs. 1(b) and 1(c). The imaging system frame was constructed from extruded T-slotted aluminum bars (80/20 Inc.) and has approximate dimensions of $2.82\text{ft}\mathrm{.}\times 3.90\text{ft}.\times 5.77\text{ft}.$ ($\text{width}\times \text{length}\times \text{height}$). Custom lead-lined stainless-steel panels and door were designed and fabricated (BFK Innovation Inc.) and were mounted to the frame to protect from x-ray leakage as well as allowing for the system to be light-tight.

Fig. 1

The FXLT Imaging System. (a) Proposed scanning scheme for the FXLT system; (b) CAD model of the proposed system; and (c) physical build of the scanner.

3.

X-ray Excitable Contrast Agents

We will employ rare earth-doped nanophosphors as contrast agents in this multiplexed bioimaging system. For example, ${\mathrm{NaGdF}}_4:\mathrm{Eu}$ or Tb was synthesized using a citrate method, involving lanthanides-citrate complex formation followed by nucleation and growth upon the addition of NaF. Synthesized nanoparticles were washed and finally dispersed in aqueous solution (pH 7). Following the synthesis, the nanoparticles were annealed at a high temperature (above 850°C) to increase the emission intensity by removing defects that can cause luminescence quenching. To prevent particles sintering during this annealing, ${\mathrm{NaGdF}}_4$ nanoparticles were encapsulated within a silica shell using the Stöber process, as described below. The synthesized nanoparticles were characterized using dynamic light scattering (DLS), transmission electron microscopy (TEM), and x-ray excited optical luminescence (XEOL) spectroscopy (Fig. 2).

Fig. 2

(a) Luminescence spectra of ${\mathrm{NaGdF}}_4:\mathrm{Eu}$ (red) and ${\mathrm{NaGdF}}_4:\mathrm{Tb}$ (green) nanophosphors, TEM images above the spectra show silica coated particles. (b) TEM image of biotin-coated nanophosphors (top), and STEM image of biotin-coated nanophosphors adhered to streptavidin coated microspheres (bottom), inset image is a zoomed-in view.

Luminescent ${\mathrm{NaGdF}}_4$ nanoparticles were encapsulated with silica layers via the Stöber process, modified with (3-aminopropyl) triethoxysilane, and lastly functionalized with carboxylic acid–PEG–Biotin (1000 Da) using EDC coupling chemistry in aqueous MES buffer. The surface functionalization with biotin was initially verified by scanning electron microscopy imaging, which showed that biotinylated nanophosphors adhered to streptavidin-coated silica microspheres. Surface charge and size of the biotinylated nanophosphors were measured using DLS.

Synthesized nanoparticles emit visible light photons when irradiated with x-rays. Both DLS measurement and TEM indicate that the size of the nanophosphors without any modifications is around 100 nm. According to XEOL spectroscopy measurements, both Eu- and Tb-doped ${\mathrm{NaGdF}}_4$ nanophosphors show relatively low emission intensity, likely due to lack of luminescent centers or self-quenching, respectively. Later, hydrothermal treatment increased the emission intensity by a factor of 2 to 3 (annealing without a silica shell increased the emission intensity by another factor of 5 but resulted in sintered particles, negating their biological application). After annealing silica-coated ${\mathrm{NaGdF}}_4$ particles, there was minimal increase in intensity compared to pristine nanoparticles. Later, silica shelled nanophosphors were functionalized with biotin-labeled PEG molecules. Post functionalization steps, the nanophosphors bind well with a streptavidin-coated glass beads (Akadeum Life Sciences) demonstrating functionalization. Additional information and comparisons using the nanophosphors can be found in Ref. 23.

4.

Machine Learning-Based XLCT Reconstruction Algorithm

4.4.

Validation of Deep-learning Method

We performed testing to demonstrate the proposed deep learning-based image reconstruction method based on 3D digital cylindrical phantoms. The optical parameters of the cylindrical phantoms were set to around $15{\mathrm{cm}}^{-1}$ for the scattering coefficient and around $0.5{\mathrm{cm}}^{-1}$ for the absorption coefficient. Four target regions were implanted in the phantom. The phantoms were irradiated by an x-ray beam to excite nanophosphors to emit luminescent light, and three optical fiber bundles equiangular distributed upper surface of the phantom to collect the luminescence. The focused x-ray beam has a diameter of $50\mu \mathrm{m}$, which was used for the simulation of x-ray imaging. The x-ray beam was translated with an increment of 0.5 mm in $X\text{-}Y$ plane for 50 times to scan the phantom for each projection view. Total 15 projection views were acquired to generate sinogram for the tomographic imaging. Then the sinogram with 15 views was input to the trained ResNet model to synthesize new sinogram with 30 projection views. Then, the iterative image reconstruction algorithm with the minimization of the image total variation was applied to reconstruct nanophosphors concentration distribution from the synthesis sinogram. The nanophosphors distribution generates clearly patterns in sinogram, which is very effective to perform the interpolation task using deep learning techniques. Our results have shown that the proposed ResNet can effectively reduce noise and artifacts, and improve spatial resolution images, as shown in Fig. 3. The peak signal-to-noise ratio (PSNR) and SSIM are utilized to quantitatively evaluate the performance of the deep learning image processing methods, achieving PSNR of 24.09 and SSIM of 0.7993 for representative slices, whereas the image reconstructed with few-view projection data has PSNR of 19.48 and SSIM of 0.6968.

Fig. 3

Projection data synthesis and image reconstruction. (a) Sinogram of 15 projection views; (b) sinogram of 30 projection views; (c) reconstructed sinogram from 15 projection views; (d) image reconstructed from 15 projection views; (e) ground truth image of phantom at a plane; and (f) image reconstructed from reconstructed 30 projection views.

5.

Conclusions and Future Works

XLCT has become an attractive imaging modality for the 3D high-spatial resolution imaging of deeply embedded x-ray excitable contrast agents with good measurement sensitivity. Here, we have first shown the design and build of our small-animal dedicated FXLT scanner that can perform both $\mu \mathrm{CT}$ imaging and 3D high-resolution XLCT imaging in one system. Having both the ability to perform the $\mu \mathrm{CT}$ imaging immediately before the XLCT imaging inside a single scanner is highly beneficial and allows for easier registration of the anatomical and optical imaging since the object orientations are guaranteed to be the same. With our previous setup, having to move the object to two different scanners for the imaging made it harder to ensure both CT and XLCT were scanned in the same orientations.^5–7 A few other additions to the FXLT scanner will also allow for improved imaging performance compared with our prototype system in Ref. 7. First, the new scanner’s x-ray beam diameter is much smaller than the previous scanner (49.9 versus $101.5\mu \mathrm{m}$), which will allow for higher spatial resolution capabilities. From our previous reports,^5,27,28 we can estimate the spatial resolution of the designed FXLT scanner to be about $94\mu \mathrm{m}$. Next, the use of four highly sensitive PMTs (H7422-50) to detect the emitted optical signals will allow for much more sensitive setup than previously where only one PMT was used. In addition, we have recently switched from silica core optical fiber bundles to liquid-light guides, which allow for higher optical transmission efficiency, especially in the NIR range, which means we expect much more signal to be delivered to each PMT than previously. From our previous study regarding the sensitivity of XLCT for ${\mathrm{Gd}}_2{\mathrm{O}}_2\mathrm{S}:{\mathrm{Eu}}^{3+}$, we estimated that the limits of detection (LOD) is approximately $2\mu \mathrm{g}/\mathrm{ml}$.²⁹ With this new setup, we expect to achieve an even lower LOD, which will be verified in future studies. With additional PMTs, we are also capable of performing imaging of different nanoparticles with different emission wavelengths as well, which can be useful for co-registered imaging of multiple targets.

We have also synthesized 100 nm ${\mathrm{NaGdF}}_4$ (both Eu and Tb doped) nanophosphors with a silica coating and functionalized with biotin, which were successfully adhered to streptavidin-coated microspheres (Fig. 2). These particles have shown distinct luminescence spectra compared to previous ${\mathrm{Gd}}_2{\mathrm{O}}_2\mathrm{S}:{\mathrm{Eu}}^{3+}$ particles,³ which will allow for multi-color XLCT imaging. Lastly, we have also proposed and developed a deep-learning-based reconstruction algorithm for XLCT imaging based on the ResNet. In our preliminary work shown in this study, we can see a clear enhancement in the XLCT reconstruction image quality when comparing the few-view dataset image reconstruction [Fig. 3(d)] to the reconstructed full-view dataset [Fig. 3(f)] with our method.

Overall, in this work, we proposed, designed, and built the first reported small-animal dedicated FXLT scanner using a rotary gantry, synthesized ${\mathrm{NaGdF}}_4$ nanophosphors with unique emission spectra. We proposed a deep-learning-based XLCT reconstruction algorithm and have shown our preliminary results from these three aspects in this study. Our next direction with the FXLT scanner is to develop a hardware control and image acquisition software for performing both the $\mu \mathrm{CT}$ and FXLT scans. Future work with the nanophosphors involves optimizing synthesis, annealing, and surface modification protocols to obtain brighter nanophosphors and mixing with ${\mathrm{Gd}}_2{\mathrm{O}}_2\mathrm{S}:{\mathrm{Eu}}^{3+}$ nanophosphors for multicolor imaging. We will plan to first test and compare the already synthesized nanophosphors as potential XLCT imaging contrast agents by performing preliminary XLCT scanning using our previously developed system in Ref. 7. Later, we can develop nanospheres with brighter luminescence and longer stability, and their application will be tested as in situ light sources for detecting target receptor molecules in vivo. For the proposed deep-learning-based reconstruction algorithm, further testing will be performed to further verify our proposed reconstruction method particularly using datasets acquired from physical experiments. Finally, the ultimate goal will be to synergistically combine all aspects of this study (the FXLT scanner, synthesized nanophosphors, and developed reconstruction algorithm) by performing XLCT scanning of a mouse with our FXLT scanner, using developed nanophosphors as contrast agents, and then using our proposed algorithm to reconstruct the FXLT images.