Objective. Magnetic nanoparticles can be used as a targeted delivery vehicle for genetic therapies. Understanding how they can be manipulated within the complex environment of live airways is key to their application to cystic fibrosis and other respiratory diseases. Approach. Dark-field x-ray imaging provides sensitivity to scattering information, and allows the presence of structures smaller than the detector pixel size to be detected. In this study, ultra-fast directional dark-field synchrotron x-ray imaging was utlilised to understand how magnetic nanoparticles move within a live, anaesthetised, rat airway under the influence of static and moving magnetic fields. Main results. Magnetic nanoparticles emerging from an indwelling tracheal cannula were detectable during delivery, with dark-field imaging increasing the signal-to-noise ratio of this event by 3.5 times compared to the x-ray transmission signal. Particle movement as well as particle retention was evident. Dynamic magnetic fields could manipulate the magnetic particles in situ. Significance. This is the first evidence of the effectiveness of in vivo dark-field imaging operating at these spatial and temporal resolutions, used to detect magnetic nanoparticles. These findings provide the basis for further development toward the effective use of magnetic nanoparticles, and advance their potential as an effective delivery vehicle for genetic agents in the airways of live organisms.

Cystic Fibrosis is caused by pathogenic variants of the cystic fibrosis transmembrane conductance regulator (CFTR) gene. This gene encodes for the CFTR protein, which controls the balance of chloride and bicarbonate ions within airway cells. Defects in the CFTR protein, or the absence of CFTR protein, lead to symptoms in multiple organs, some life-threatening. Within the respiratory system, cystic fibrosis dehydrates the airway surface and produces thickened mucous, allowing a cycle of infection and inflammation to initiate. This leads to lung disease, which is the leading cause of morbidity and mortality amongst those living with cystic fibrosis (Vallieres and Elborn 2014). Recently, CFTR modulator therapies have been shown to be effective in restoring ion channel function, reducing the severity of symptoms for people with some classes of CFTR mutation (e.g. elexacaftor/tezacaftor/ivacaftor for class 2 mutations) (Middleton et al 2019). However, for people with class 1 CFTR mutations (those in which no CFTR protein is produced) these drugs are not effective, leaving them without an effective treatment option (Lee et al 2021).

Another proposed therapy is gene-addition therapy, whereby a fully functioning copy of the CFTR gene is inserted into a cell by a transfer vector (Lee et al 2021). This copy of the gene produces functional CFTR protein, and so should restore airway hydration and mucous clearance. However there are some challenges with gene therapy, as discussed by Donnelley and Parsons (2018). Gene vectors need to be transported to the specific airway cells that require correction, and retained in place long enough to effectively interact with the target cells, a factor termed residence time. Achieving sufficient residence time is made challenging by mucociliary clearance, the process whereby cilia clear inhaled particles up the airway towards the mouth. A novel approach to this challenge is coupling the viral vector to magnetic nanoparticles, allowing the location of the vector to be controlled within the airways by an external magnetic field (Dames et al 2007, Alvizo-Baez et al 2016). A review on the topic with a focus on cystic fibrosis by Tan et al (2020) has detailed the process and provided information on the choices of gene vector and magnetic nanoparticles.

Controlling and visualising the magnetic nanoparticles within the airways are challenges. Their motion within pulsatile blood vessel fluid can be modelled (Heidsieck et al 2012), but the non-stop, complex motion of the reversing air (breathing) and liquid (mucociliary clearance (Gardner et al 2020)) environment within the airway surface extends the challenges further. Simply visualising the magnetic nanoparticles to develop control mechanisms is difficult, however Donnelley et al (2022) have had some success using high-resolution x-ray imaging. That study utilised propagation-based phase-contrast effects to enhance image contrast, and showed that strings of magnetic nanoparticles formed in the airways of rats when a static external magnetic field was applied, similar to the strings that iron filings form under the influence of magnet fields. In the present study we have for the first time explored the use of dark-field x-ray imaging to further enhance nanoparticle contrast and gain new insights into magnetic particle control in vivo.

Dark-field imaging detects ultra-small angle scattering caused by microstructures within the sample (Pfeiffer et al 2008). If these structures are asymmetric and the longest axes are aligned, they can scatter anisotropically (i.e. x-rays are scattered preferentially in one direction, usually orthogonal to the orientation of the long axes of the scattering object). Detecting this scatter gives the directional dark-field signal (Jensen et al 2010), allowing the orientation of microstructures to be measured. Previous biological applications of dark-field imaging have focused heavily on lung imaging (Schleede et al 2012, Willer et al 2021), as the alveolar structure produces a strong dark-field signal (Bech et al 2013), and dark-field imaging has also been used in clinical lung imaging (Viermetz et al 2022). Microstructures in other organs such as the monosodium urate crystals that cause joint issues in gout have also been visualised using dark-field imaging (Braig et al 2019).

The above examples all relied on having a static sample, as the techniques used to extract the dark-field images typically require multiple exposures; single-shot techniques (Zanette et al 2014, Vittoria et al 2015) avoid this. Doherty et al (2023) used single-shot imaging with a laser-driven x-ray source to achieve femtosecond speed dark-field imaging with a beam-tracking technique. Massimi et al (2021) developed a method for imaging dynamic samples that change slowly (relative to the exposure time) using a similar beam-tracking technique, but using multiple exposures to increase spatial resolution. That approach is suitable for slower transitions such as the change in the structure of melting metal (Massimi et al 2022), but is impractical for imaging airways in live anaesthetised rats, since the airways are moving rapidly compared to the exposure time we use. Gating the motion of a grating to the motion of a ventilator allowed Gradl et al (2019) to record dark-field lung images across the breath cycle in mice with a grating interferometer. However, this approach requires a repeatable lung motion and lacks the spatial resolution required for our study.

The one-dimensional optics used for these dynamic examples allowed only for scalar dark-field information (i.e. without any orientation aspect) to be extracted. In contrast, two-dimensional optics, such as circular gratings also allow for single-shot directional dark-field imaging (Kagias et al 2016). Their use has also been expanded to fast tensor tomography (Kim et al 2021). Other two-dimensional optical elements such as sandpaper (Zhou et al 2018) and grid-like two-dimensional gratings (sometimes referred to as single-grid imaging) (Dreier et al 2020) can also be used for single-shot directional dark-field imaging. Dynamic directional dark-field imaging was demonstrated by Croughan et al (2023) using a single-grid beam-tracking technique (How and Morgan 2022). Bennett et al (2010) demonstrated that single-grid imaging can be used for in vivo imaging utilising phase-contrast, however their use of a laboratory x-ray source meant the exposure time (28 s) was too long for dynamic biological imaging.

The aim of this study was to explore how magnetic nanoparticles used as a delivery vehicle within a live airway behave under the influence of dynamic magnetic fields. We hypothesised that the magnetic nanoparticles would produce a dark-field signal, and that when they formed 'strings in a magnetic field, those strings would produce a directional dark-field signal. A directional dark-field signal in this context could both assist with visualising the presence of the nanoparticles amongst a background that lacks any orientation, and could reveal how the particles are moving in the presence of the magnetic field.

2.1. Synchrotron setup

The experiment was undertaken at the BL20XU beamline of the SPring-8 Synchrotron, where the experimental hutch is 245 m from the electron storage ring. A 25.0 keV monochromatic beam was used. An Orca Flash 4.0 (Hamamatsu, Japan) sCMOS detector (2048 × 2048 pixels) was coupled to a 10 μm GAGG:Ce scintillator via a microscope lens giving pixel size in the sample plane of approximately 0.51 μm. A fast x-ray shutter (Uniblitz XRS6, Vincent Associates) was used to minimise radiation dose.

2.2. Animal preparation and magnetic control

The rat was anaesthetised with a mix of medetomidine, midazolam and butorphanol for the duration of the experiment, and kept warm with a heat lamp during imaging. As per recommendations in Morgan et al (2020), fur in the region of interest was removed as it can induce artefacts. A 16 G a i.v. cannula (BD Insyte) was inserted into the trachea as an endotracheal tube.

A custom 3D-printed motorised magnet holder, shown in figure 1(b) and based around a lazy-susan bearing (Bunnings, Australia), was designed to externally rotate a magnet around the rat neck and chest. The magnet holder was configured to contain a single 19 mm diameter (28 mm length) nickel-cased rare-earth neodymium iron boron (NdFeB) magnet (N35, cat # LM1652, Jaycar Electronics, Australia) with a remanent magnification of 1.17 Tesla. The magnet rotation rate was varied using a pulse-width modulation motor controller (RS Components, Australia).

Figure 1. (a) Shows a simplified top-down view of the experiment. The beam is patterned by the grid, creating beamlets. These beamlets are diffused as they scatter through the sample, with the dark-field extracted from the strength of the blurring seen downstream at the detector. Note how the trachea and centre of rotation of the magnet are aligned at approximately 45° to the beam. A photo of the motorised magnetic array and rat are shown in (b). The red glow is from the heat lamp, used to maintain animal temperature. (c) Shows how the blurring of the beamlets is modelled by the UMPA algorithm, using an intensity scaling for transmission T, and convolution with a directional dark-field kernel, where that kernel is parameterised by major and minor axis widths σmajor and σminor and rotation angle θ. (d) shows that magnetic nanoparticles in a capillary tube (shown in transmission in the upper image) produce a directional dark-field signal (lower image).

Download figure:

Standard image High-resolution image

The rat was placed supine within the centre of the device, and positioned so the beam passed through the trachea at approximately 45°, as can be seen in the top down view in figure 1(a). A small polyethylene delivery cannula was inserted into the endotracheal tube to reach the region of interest within the trachea. The diameter difference between the cannula and the delivery tube was sufficiently large that the rat could freely breath without the need for mechanical ventilation. A humidified air supply was used to ensure the trachea remained well hydrated, as the beamline environment contains dehumidified air.

Silica superparamagnetic nanoparticles of nominal size 0.95–1.4 μm (Spherotech, USA, serial number SIM-10-10H) at 2.5% concentration in distilled water were used. Methylcellulose of 0.25% was added to the solution to increase the viscosity. No viral vector was bound to these nanoparticles, as the purpose of the experiment was only to determine our ability to visualise and control the magnetic nanoparticles. Particles were injected into the trachea via the polyethylene cannula using a remotely controlled syringe pump (WPI UMP3 UltraMicroPump), at a flow rate of 1 μl s−1.

Inspiration was detected using a laser sensor (Donnelley et al 2023), and the camera was triggered at the same point of each breath of the rat, with eight 0.45 ms exposures triggered every time the rat reached peak inspiration. The rat was humanely killed at the end of the study.

2.3. Dark-field imaging

To extract the x-ray dark-field signal, a checkerboard π-shifting phase grid with 14.4 μm period was placed into the beam 63 cm upstream of the sample, to produce a grid pattern with 7.2 μm period at the detector. The grating pattern is 32 μm deep, on a 300 μm thick silicon wafer, giving 86% transmission. The manufacturing process is described by Rutishauser et al (2013) and the first use of this particular grating reported in Morgan et al (2013). As seen in figure 1, the small pixel size allowed a finer pattern to be seen within each grid hole, at least in the 'Grid only' image. The sample was 44 cm upstream of the detector. Figure 1(a) shows a simplified diagram of the grid producing a pattern of beamlets, with the sample scattering each beamlet, blurring the image of the beamlets seen at the detector. 100 reference frames (images without the sample) were taken, as well as flat-field and dark-current images for beam and detector correction.

The dark-field images were extracted using the directional dark-field extension to the unified modulation pattern analysis (UMPA) algorithm, presented by Smith et al (2022) and based on the works by Zdora et al (2017) and De Marco et al (2023). UMPA is a versatile algorithm, proven to work with a variety of x-ray sources (Zdora et al 2020, Smith et al 2023 ,Smith 2023), and previously shown to work with periodic patterns such as those used in single-grid imaging (Gustschin et al 2021, Riedel et al 2021).

The algorithm works by comparing windows in the images the pattern produces with and without the sample in the beam. Attenuation in the sample is measured by looking at the reduction of intensity in the pattern, and phase can be found by looking at how the beamlets are refracted (i.e. how the pattern at the detector is translated). The directional dark-field signal is modelled as the blurring of the pattern by a two-dimensional Gaussian kernel. This Gaussian is defined by the major and minor components, σmajor and σminor , as well as the orientation θ. A graphic representation of this is shown in figure 1(c).

An analysis window of 41 × 41 pixels was used, as this gave a good signal to noise ratio (but at the cost of spatial resolution). This is a larger window than typically used for phase-contrast applications, but Savatović et al (2024) showed windows this large can provide good results for dark-field imaging. This single-sample-exposure setup allows us to image at the same frame rates as would be possible for simple attenuation or propagation-based imaging. However, it must be noted that applying a single-sample-exposure dark-field extraction algorithm will reduce the spatial resolution of the transmission image extracted when compared to the transmission image that would be measured without the dark-field setup. This is because multiple quantities (T, σmajor, σminor, θ, phase) must be extracted at a given location in the image, and hence multiple measurements must be taken, in this case spatially separated over multiple adjacent pixels. In the case that a higher spatial resolution was required, the measurements would need to be temporally separated over multiple exposures captured for different reference pattern positions (Smith et al 2022). A multiple-exposure approach would be incompatible with the movement seen exposure-to-exposure with this small-area live animal imaging.

Images were taken of 'strings' of magnetic nanoparticles in a capillary tube to confirm that they produced a directional dark-field signal, with transmission and directional dark-field images shown in figure 1(d). As is described in Smith et al (2022), the directional dark-field image uses a HSV colour scheme. The hue (H) relates to the scattering orientation θ and the saturation (S) to the eccentricity of the Gaussian kernel. The value (V) is the scattering magnitude, defined as . σmajor and σminor (and the magnitude calculated from them) are in pixel units, and multiplying these values by converts them to the angle of the scattering cone of the beam. A copy of the data is available at Smith 2024 and can be analysed using the code at Smith 2022 and De Marco 2022

3.1. Magnetic nanoparticle delivery detection

The delivery of the magnetic particles into the trachea must be performed successfully to allow them to be later manipulated with the external magnet. No magnetic field was present during the delivery, and thus we expected that the magnetic particles would be evenly dispersed throughout the saline solution used to deliver them. In this situation an isotropic dark-field signal was expected, which one would assume does not require directional dark-field imaging to visualise. However, there were a number of other features within the image with a strong directional dark-field signal caused by edge scattering (i.e. a dark-field signal seen along the edge of a sample), such as the edge of the delivery tube and wrinkles in the rat's skin, visible in figures 2(a)–(d). This edge scattering was highly directional, with a large dark-field blur orthogonal to the edge at which it occurs, but with much smaller scattering in other directions.

Figure 2. Transmission (a), directional dark-field (b), major (c) and minor (d) components of the dark field images highlight different features within the airway during delivery. Anisotropically scattering features are less prominent in the minor dark-field component, increasing contrast of the isotropically scattering magnetic nano-particles. Note that (c) and (d) use a different colour scale. (e) Shows multiple frames of the delivery for transmission (above) and minor dark-field component (below). (f) Shows the mean signal values across the sequence within the ROIs shown in (a) and (d).

Download figure:

Standard image High-resolution image

The suspension of magnetic nanoparticles should scatter isotropically, so the major and minor scattering components (σmajor and σminor as described above) should be equal. For the background features, the major scattering component is large, but the minor scattering component is much smaller. Thus, in this case, focusing on the minor scattering component gives a stronger signal from the magnetic nanoparticles compared to the background signal, as is clear from figures 2(c) and (d). Figure 2(e) shows a selection of frames taken from a sequence during the delivery of particles. 25 μl of a solution containing particles were delivered over 25 s, with 180 frames captured during this period. Note that the frames were not evenly spaced in time as the camera was triggered based on the rats breathing, which naturally varied slightly, so times are approximate. The sequence begins with a small bubble of air being expelled from the end of delivery tube (t = 0 s). Liquid then begins to flow, creating the diagonal line as it flows from the tube onto the bottom surface of the trachea. Between t ≈ 0.42 s and 22.35 s there was no visible change detected in either the transmission or dark-field signals. At t ≈ 22.91 s there was a noticeable change in the dark-field signal in the lower half of the tube, and the feature became more clear in subsequent frames. The end of the sequence saw this feature being pushed along by the air bubble at the end of the delivery.

Some change was also visible in the transmission signal in figure 2(e). In order to compare the two, mean values for transmission and dark-field magnitude were calculated within a 100 × 200 pixel region located in the centre of the tube, the location is shown in figures 2(a) and (d). These values are plotted in figure 2(f). The mean background signal strengths (i.e. the signal seen in the absence of particles) were calculated by taking the average value between t ≈ 6.9 s and 20.8 s (note that the figure and calculations do not begin at t = 0 s as the animal moved slightly, moving the ROI). The standard deviation across the interval was used as an estimate for the noise. The average values for transmission and dark-field within the region both peaked at t ≈ 24.02 s. The strongest attenuation was seen as a transmission value that was 5.2 standard deviations below the mean background value (figure 2(f)). The dark-field signal peaked at 18.6 standard deviations above the background signal, showing that dark-field increased the signal-to-noise ratio of this event 3.5 fold.

3.2. Rotating magnetic nanoparticle strings

Having seen that an external magnetic field causes the magnetic nanoparticles to create packed string-like structures (see methods), we rotated a magnet around the rat's trachea to cause these strings to form and to rotate. To explore if this rotating field could move delivered particles within the airway, we examined whether directional dark-field imaging could be used to show the change in orientation of the particle strings. Figure 3 shows four time points as the magnetic field passes over a region of magnetic particles that had deposited in the endotracheal tube used to guide the delivery cannula into the trachea. The transmission images (left column) show the strings of particles are initially aligned vertically as the magnet is above the rat, and stay aligned to the magnetic field as the magnet moves. Due to the way the magnet and rat are oriented to the beam, as the magnet continues along its 45° offset-circular path around the trachea, there is a component of the magnetic field lines that runs parallel to the beam (i.e. into the image). This creates strings with a component along the imaging axis, leading to the pattern seen in the second and third rows.

Figure 3. X-ray transmission (left), directional dark-field (middle), and magnitude of the directional dark-field signal (right) images taken from a time series as the magnet passes around the rat (magnet symbol shows the approximate direction of magnetic field). Magnetic particles within the delivery tube form ordered adjacent strings against the top of the tube within the airway, stretching roughly halfway along the y-axis in the field of view. The colour map for the directional dark-field plot uses a hue—saturation—value (HSV) scheme. The hue indicates the orientation of the detected scattering. The saturation shows how directional or asymmetric it is, with white/grey pixels indicating isotropic scattering occurred. The magnitude, or overall scattering strength, is given by the value channel.

Download figure:

Standard image High-resolution image

The middle column shows the directional dark-field images. In the middle two rows, the strings of magnetic particles produce a directional dark-field signal that can be easily seen by eye. The orientation of the measured scattering is around 45° (blue) and −45° (yellow–green). However, the signal is not clearly visible in the top and bottom images, where the particles are aligned vertically. The lack of clear directional dark-field signal in these images is confirmed by looking at the dark-field magnitude in the third column.

We believe that the dark-field signal is only visible in the middle two images of figure 3 because the strings of magnetic particles are partially aligned along the axis of the beam when the magnet passes alongside the rat. This means the beam passes through many more magnetic particles compared to the images where the particle strings are aligned only in the vertical direction, leading to a stronger scattering of the beam and a stronger dark-field signal.

The ability to visualise the magnetic nanoparticles via their dark-field signal shows a significant advantage over the phase contrast imaging used by Donnelley et al (2022). The delivery image sequence shows that the magnetic particles are unexpectedly agglomerating or settling within the delivery tube for some reason, with the final 10% of the delivery period containing a substantially greater number of magnetic nanoparticles. The reason for this non-uniformity during delivery is not known, we speculate that gravity may cause the magnetic nanoparticles to settle, with the delivery liquid flowing past them, until the bubble present at the end of the delivery forces the magnetic nanoparticles and remaining liquid from the tube. This non-homogeneous particle was not previously visible without dark-field imaging, and could prove detrimental to cellular uptake of viral vectors if all the magnetic nanoparticles are delivered to only a small area.

As the experiments occurred inside the trachea of a live anaesthetised and spontaneously breathing rat, there was a background dark-field signal caused by tissue, skin, and residual hair. Previous demonstrations of directional dark-field imaging using similar hardware by Croughan et al (2023) and a similar analysis technique by Smith et al (2022) did not have a background signal to interfere with image clarity. In the delivery experiment, we were able to overcome this limitation by utilising the minor component of the directional dark-field signal, so the largely directional background signals caused by edge scattering were suppressed. This approach is not possible using only a scalar dark-field extraction algorithm. In non-living systems, enhancing contrast by subtracting the background signal (taken by imaging prior to the event of interest) is possible, but would not be practical here. Image registration would be unreliable due to small shifts through the depth of the rat's neck and trachea, and since the rat's breaths are not identical, even when images are taken at identical points in the breath cycle, the position of background features is subtly different. There are also several distinct motions (the rat moving during breathing, the smaller background pulsing from the heartbeat, as well as the motion of the magnetic nanoparticles themselves) that make any technique that requires multiple exposures (such as Massimi et al (2021)) impractical for live animal airway imaging.

The analysis window used by the UMPA algorithm (Smith et al 2022) is a Hamming Window (Blackman and Tukey 1958) rather than the uniform window shown for simplicity in figure 1(c). This makes calculating the spatial resolution of the system challenging, as information from across 41 × 41 0.51 μm pixels (so 22 μm×22 μm) is used to give the reconstruction, but the pixels are not weighted evenly. Taking the full-width at half-maximum of the Hamming window as an estimate for the spatial resolution gives 11 μm. In an attempt to increase the spatial resolution, a finer phase grating that produced a pattern with period 2.7 μm was tested. It showed good results when imaging the magnetic nanoparticles in a capillary tube, however extracting any dark-field signal from the in vivo experiments was not possible. We believe the smaller features in the reference pattern were more vulnerable to noise and additional similar-scale contrast introduced by the sample.

Some optimisation of the imaging procedure could improve these dark field imaging capabilities. Increasing the sensitivity of the technique by increasing the sample-to-detector distance would allow the dark-field blur induced to become more easily detected. Increasing the beam flux or detector efficiency would allow for either a reduced photon noise with the same exposure, or a reduced exposure time, potentially reducing any motion blur. Using the single-grid imaging presented here with the laser-driven sources presented by Doherty et al (2023) could extend their technique to provide directional dark-field imaging with a femtosecond temporal resolution, but it is not clear if the x-ray beam would have sufficient flux to penetrate the neck of a live rat sufficiently to produce useable images, given the lower peak energy. As we used the UMPA algorithm to extract the dark-field image, quantitative phase-contrast images could also be extracted (Zdora et al 2017, De Marco et al 2023), giving a third contrast mechanism. However, this approach remains to be examined.

Dose rates were not measured during the experiment, however at the same synchrotron beamline using a similar setup, Donnelley et al (2014) calculated the dose for a 50 ms exposure as approximately 46 mGy. Assuming the beamline and synchrotron configuration is the same and ignoring the effect of the grating, this gives an estimated approximate dose of 43 mGy per image with our exposure time of 45 ms. Further dose measurements may be important if recovery experiments were to be considered in future. The patterned beam bears resemblance to those being investigated for synchrotron-based microbeam radiation therapy (Engels et al 2020), which show a complex relationship between cell health and microbeam radiation. We speculate wildly that the patterned beam developed for microbeam radiation therapy could potentially be used for dark-field imaging of a tumour during the therapy. The single-shot directional dark-field demonstration by Dreier et al (2020) showed in vivo technique should be compatible with a laboratory source if the speed needed to capture high-resolution images within the airways was not a consideration. However, it is not clear that the dose would allow for recovery.

Our intention was to use this technique to detect the presence of particles in areas where the concentration was lower and the particles were invisible in transmission by detecting a directional dark-field signal rotating in concert with the magnet. Although we have shown proof-of-convept for in vivo directional dark-field airway imaging, we expect that further optimisation of the imaging setup will allow us to improve the feasibility of using magnetic nanoparticles as a gene vector delivery mechanism. Propagation-based phase contrast x-ray imaging has been used to show that the speed at which mucociliary transit moves particles within the trachea changes depending on the coating of the particles (Gardner et al 2019). Dark-field imaging could build on this work, potentially enabling quantitative comparisons of how particle properties affect particle motion and deposition.

This experiment demonstrated that single-shot directional x-ray dark-field imaging is a feasible approach in a complex in vivo imaging environment, and that dynamic events could be captured. The fastest possible acquisition speed using this technique is identical to that of the imaging setup in a 'standard imaging mode, when no dark-field is considered; the only cost to the use of dark-field imaging being a loss in spatial resolution in the retrieved images. Here, dark-field imaging allowed the agglomeration of magnetic nanoparticles during the delivery to the trachea to be detected with a 3.5 times increase in signal-to-noise ratio compared to the transmission signal. The ability to visualise these small particles more easily with dark-field will allow visualisation of lower particle concentrations in future, and with further optimisation of the delivery and imaging setups and refinement of the techniques described, there is opportunity to expand the in vivo detectability of delivered therapeutic particles.

We would like to thank Michelle Croughan (Monash University, Australia) for her assistance, discussing algorithms that can be used to extract the directional dark-field signal.

The synchrotron radiation experiments were performed at BL20XU of SPring-8 with the approval of the Japan Synchrotron Radiation Research Institute (JASRI) (Techniques developed across proposals 2022A1077, 2022B1169, 2023A1212, 2023B1272). We acknowledge travel funding provided by the International Synchrotron Access Program (ISAP) managed by the Australian Synchrotron, part of ANSTO, and funded by the Australian Government.

Studies were supported by the Medical Research Future Fund (RFRHPSI000013), Cystic Fibrosis Foundation Grant (DONNEL21GO), and Cystic Fibrosis Australia's Geelong Genetic Therapies Innovation Grant. KSM acknowledges funding from the Australian Research Council (FT18010037).

This work was supported with supercomputing resources provided by the Phoenix HPC service at the University of Adelaide.

The data that support the findings of this study are openly available at the following URL/DOI: https://doi.org/10.25909/24085896.

Removed for double-anonymous reviewing. All animal studies were performed in accordance with protocols approved by the University of Adelaide (M-2020-022) and SPring-8 Synchrotron animal ethics committees.

CRediT Contributions. RS: Conceptualization, Methodology, Formal analysis, Investigation, Data Curation, Writing—Original Draft, Visualization. KM: Conceptualization, Methodology, Investigation, Writing—Review and Editing. AM: Investigation, Writing—Review and Editing, Funding acquisition. PC and NR: Investigation, Writing—Review and Editing. DP and MD: Conceptualization, Investigation, Writing—Review and Editing, Resources, Funding acquisition, Supervision.