In this exploratory, proof-of-mechanism, ex vivo study, we have shown that a microbubble contrast agent can give contrast enhancement in clinically compatible GIM. Both the iodine and the microbubbles were visible in GIM, but detectable with two different image signals. This illustrates one of the key features of the GIM technique, namely, that it provides a set of three different images, each based on different physical properties, hence, potentially carrying different information: iodine is visible in the attenuation image because of its high attenuation coefficient while microbubbles become visible in the DF image due to its generation of ultra-small-angle scattering. Since microbubbles do not attenuate x-rays, the attenuation image can be used for anatomical orientation and the corresponding DF image for the assessment of contrast enhancement, which would result in a contrast-enhanced mammogram at a lower radiation dose and with a shorter acquisition time compared to conventional contrast-enhanced mammography [26]. Furthermore, microbubbles have different contrast kinetics compared to gadolinium and iodine than can be exploited in GIM. The assessment of wash-in and wash-out curves in the differentiation between benign and malignant lesions is a valuable tool in breast MRI, but the same has not been proven for iodine contrast-enhanced mammography [27]. As opposed to an early gadolinium wash-out signifying malignancy, microbubbles have been shown to have a persistent enhancement in malignant lesions [28], making it suitable for a bilateral mammography procedure. Finally, since microbubbles can be loaded with, e.g. drugs or genes, in combination with unloaded bubbles for imaging, there is a potential use also in theranostics [29, 30].
We found that the mean CNRs of the microbubbles were lower to that of iodine, which was expected especially considering that both contrast agents were used in its undiluted form. Iodine is normally diluted in clinical practice, whereas microbubbles are not. The main aim of this study was, however, only to determine whether there was a visible enhancement with the microbubbles in DF mode or not. Concerning test reliability, the experiment was repeated four times, and in all cases, the microbubbles were visible in the dark-field mode. However, there were several factors that led to a variation in the measurements such as the force applied to the contrast injection and a variation in sample thickness. Most importantly, the samples were not homogeneous leading to an uneven infiltration of the directly injected contrast agents.
If microbubbles are to be used as a contrast agent in GIM in vivo, further considerations have to be made concerning dose, and most importantly, the proper size of the bubbles. Theoretically, since the DF signal in grating interferometry is usually generated from small- or ultra-small-angle scattering, bubbles whose diameter matches the auto-correlation length of the setup could optimise the scattering signal [31, 32]. Consequently, the optimal bubble size for our investigational device would be 1 μm, i.e. slightly smaller than the commercial contrast agent actually used in the experiment.
To the best of our knowledge, this is the first time the feasibility of using microbubbles in a clinically compatible grating-interferometry phase-contrast imaging set-up has been shown. Velroy et al. [22] found that microbubbles scatter x-rays using grating-interferometry phase-contrast imaging in an experimental setting imaging vials with microbubbles with acquisition time and dose not optimised for clinical imaging. It is expected that the CNR of the DF contrast can be higher with a preclinical system compared with a clinical system, since the latter has constraints on geometry, radiation dose, and acquisition time. Without these constraints, the reduction of photon flux caused by the gratings can be compensated in the pre-clinical set-up to reduce the noise, by increasing the exposure time or making the interferometer shorter; in addition, since it is usually not subject to very strict geometric constraints, there is more freedom to optimise the geometry in terms of grating periods and inter-grating distances in order to have the most appropriate auto-correlation length.
There have been several studies on the DF signal as a function of the particle size [33,34,35,36]. For instance, Gkoumas et al. [36] measured colloidal suspensions of SiO2 microspheres of two different diameters (1.86 μm and 7.75 μm) in glycerine with increasingly higher concentrations (5–40%) on a synchrotron grating interferometer with an autocorrelation length of about 3.0 μm. They found that the utilised grating interferometer setup was more sensitive to higher concentrations. This outcome must be validated for our case, taking into account that we are measuring microbubble solution meant to be intravenously injected and that we are using a polychromatic setup.
However, even if we have demonstrated that microbubbles can provide visible contrast enhancement with a clinically compatible GIM device, the question whether this is a viable alternative in vivo remains to be answered and constitutes the main limitation of our study at this point. Additional studies on the appropriate bubble size as well as contrast agent concentration for GIM are needed.
In conclusion, GIM is a novel breast imaging technique at the edge of being clinically implemented. The method generates a set of three images based on the attenuation, refraction, and scattering of x-rays, which opens up the use of various contrast agents. This experiment set out to investigate whether microbubbles gave contrast enhancement ex vivo using a clinically compatible GIM setup. We found that microbubbles, due to their scattering properties, gave contrast enhancement in the dark-field mode. This implies the potential use of a contrast agent with a high safety profile, and with the prospect of using theranostics, in x-ray-based breast imaging.