In this study, the optical properties and performance of CLI using 68Ga were compared to those of 18F, under clinically relevant activity levels. The radiance was linearly associated with the activity of both radionuclides (18F: R2 = 0.98; 68Ga: R2 = 0.99). The linear relationship persisted with the addition of roughly 1 mm of the tissue surrogate. 68Ga showed a 22 times stronger radiance compared to 18F in the current study. Glaser et al. simulated the fluence rates with presence of scatter and absorption, to mimic optical properties of tissue, where a higher fluence rate was observed for 68Ga [26]. Fan et al. and Beattie et al. compared fluence rates for 18F and 68Ga measured with IVIS cameras and found radiation rates of 15 and 19 times higher for 68Ga [22, 27]. This signal intensity difference between 18F and 68Ga was slightly lower than the 26 times factor the Monte-Carlo simulations found in the literature [21]. This inconsistency may be explained by certain definitions in Monte-Carlo experiments, such as the use of true point sources in MC-simulations, whereas the relatively larger volume of an Eppendorf tube will radiate from multiple angles onto the camera. Other effects may include machine differences and non-linear detector efficiencies of the different camera systems. Furthermore, in real-life experiments, a fraction of positrons will escape from the medium before emitting Cerenkov photons [12]. Still, our results underline the superiority of 68Ga for Cherenkov imaging with respect to signal yield.
Our results indicate that this specific CLI system was not homogeneous over the entire FoV, given the found CoV of 0.18 in the raw image. Visual inspection of the CL image shows a weaker signal near the edges of the uniformity phantom. Though this observation could be due to a commonly encountered phenomenon in optics known as lens vignetting, the results may also have been influenced by the relatively small size of the phantom which matches the 60 × 60 mm FoV exactly. Accordingly, the edges of the phantom are visualized as ‘darker’ due to the lack of positron emissions from the material just outside the FoV. To account for this problem, a uniformity phantom should cover the FoV with an additional margin on all sides that is greater than the positron range. The vignetting in the FoV does not hamper the clinical assessment, since the size of average prostate is small enough to fit in the CFoV [28]. Thus, it is suggested to only use the CFoV to image the specimen. The CoV of the processed image, which is part of the clinical CLI protocol, in the CFoV was 0.07 without binning. This was considered uniform enough to leave out additional post-processing steps to improve the uniformity. The experiment was only performed with 68Ga, since uniformity of the system is expected to be independent of the radionuclide. However, the larger positron range could alter the texture of the image in comparison to a shorter range.
The resolution response was determined using a glass capillary with an outer diameter of 1.1 mm. Though the camera is able to image up to 158 μm according to the specifications, a better resolution is not deemed clinically relevant, as the surgeon is not able to resect with a higher accuracy. Still, we believe that difference in resolution between 18F and 68Ga found in this study is not entirely trivial, especially when the tissue surrogate was added. At higher binning (E300B8) this difference with tissue was roughly 0.5 mm. If a PSM occurs, surgeons are only able to shave in a specific area surrounding the PSM. Though this shaving cannot be performed with submillimetre precision, the localisation of the suspected area should be as precise as possible. That being said, a PSM found on ex vivo measurement is difficult to map back in vivo; therefore, the lower resolution of 68Ga will not hamper clinical implementation. The smaller resolution of 18F found in this study complies with literature [12, 22, 29] and can be explained by the larger positron range of 68Ga compared to 18F [30]. Direct comparison of the spatial resolution is difficult, since other groups used different setups to determine the effective spatial resolution [22, 23]. Binning is an important factor that influences the spatial resolution, since data of different pixels is combined to enhance the signal and reduce the effects of noise. Although a binning of 2 reduces the spatial resolution with roughly 10% (1.08–1.16 mm, for a 60-s exposure), the gain in signal intensity is fourfold, thus justifying the use of pixel binning in a clinical setting.
To come to a clinical image acquisition protocol that could be used during prostatectomy, various acquisition times and pixel binning setting were evaluated. The most optimal setting for clinical ex vivo 68Ga research is considered 120 s and 2 × 2 binning, thus acquiring a good spatial resolution within an acceptable timeframe for intraoperative usage and sufficient sensitivity. This is a shorter exposure time and binning factor, as the default setting for 18F (E300B8) [16, 23]. Since the light yield is higher with 68Ga, the acquisition time and binning factor can be decreased without compromising the quality of the image. For clinical implementation, time is important. For rapid assessment, the operation cannot be delayed for more than 10 min. Uptake in the prostate tumour, from analysis, shows that the uptake is sufficient for CLI imaging with 68Ga-PSMA, with the required time for prostatectomy and lesion removal.
Implications for CLI during prostatectomy
68Ga-PSMA tumour uptake measurements were performed in a heterogeneous group of prostate cancer patients. Although large variations are observed in the intensity of PSMA accumulation, we have determined an average and minimal uptake to enable clinically relevant measurements. It was stated that CLI should be able to visualize an average concentration of 3.35 kBq/mL. Based on our in vitro results the detection limit and contrast for 68Ga (with and without tissue) is sufficient to detect this average tumour uptake, even with an exposure time as low as 120 s. The detectable activity concentration with 1-mm tissue asks for an injection of 2.6 MBq/kg 68Ga 45 min prior to CLI, assuming a uniform distribution and water density in the body. Nevertheless, the tumour has 100× more receptors as benign tissue [19, 20]; thus, injected dose could be lowered for CLI visualization, thereby complying with the 68Ga-PSMA guidelines for PET imaging [31]. The standard clinical injection of ~ 100 MBq would be sufficient for intraoperative application with a protocol that fits the clinical requirements. When increasing either the binning or the exposure time, potentially an even lower radioactive dosage can be used. Still, precise patient dosage will be determined in our on-going clinical feasibility study. Prior studies with 18F showed that the radiation dose to the surgeon due to the CLI procedure was 34 μSv per procedure and 2–20 μSv per scrub nurse [14, 16]. The use of 68Ga-PSMA decreases the injected dose and thus improves the radiation safety for both the patient and personnel.
The Cerenkov signal through ~ 1-mm tissue surrogate reduced to 73% of the original signal for 18F, and 62% for 68Ga according to our measurements. However, this experimental setting does not mimic the exact clinical situation as the chicken breast was not perfused and the optical properties do not comply. The signal is influenced by scattering and absorption in tissue, the attenuation found in the current study (73%), approximates the value found in literature. Theoretical calculations of 1-mm tissue, resulted in a decrease in signal intensity of 18F of 77% [32]. The same approximation was also made for higher energy nuclides like 68Ga. However, the addition of 1-mm surrogate tissue resulted in more signal decay in our experiments (62%) than expected from literature (77%). Difference in this attenuation percentage could be explained by the influence of the refractive index (η). Tissue has a higher η resulting in a higher number of Cerenkov photons produced (see Eq. 1). The corresponding higher mass density results in a higher β attenuation cross section and concomitant reduced β particle range. Increased density therefore tends to reduce Cerenkov radiation production efficiency, but for radionuclides that emit relatively low energy β’s, the increased η dominates resulting in higher light yield for higher density materials. For the high-energy β’s of 68Ga, however, the impact of η is small and the reduction in light yield due to the density effect dominates [22]. Additionally, Glaser et al. showed that a larger refractive index has more impact on the fluence rate of 18F, as 68Ga [26].
Equation 1 is the number of Cerenkov photons N emitted per distance travelled x, which is derived from the Frank–Tamm equation [21]. β = particle velocity, η = refractive index, λ = wavelength (nm), and α = fine structure constant (α ≈ 1/137).
$$ \frac{\mathrm{d}N}{\mathrm{d}x}=2\pi \alpha \left(1-\frac{1}{\beta^2{\eta}^2}\right)\left(\frac{1}{\lambda_1}-\frac{1}{\lambda_2}\right) $$
Limitations of CLI
The interference of tissue and blood reduces the obtained signal; however, these influences were not simulated in the current in vitro set up. Therefore, it is difficult to suggest a definite ex vivo acquisition protocol upon solely in vitro measurements. Cerenkov light is predominant in the blue range of the spectrum, and attenuates towards the red part of the spectrum [33]. The weight of the spectrum changes with the influence of tissue, since the blue part is more strongly absorbed [10] as haemoglobin absorbs mostly in this part of the spectrum. For margin assessment accuracy in the penetration, depth is important, although not considered in this study. To estimate the penetration depth, it is important to have a phantom with the optical properties of the prostate, since it is influenced by scattering and absorption. The development of a prostate-like-phantom was outside the scope of this paper. For clinical application, the use of numerous filters is recommended, since it enables the possibility to only obtain emitted photons generated near the surface of the tissue. Filters would be needed to determine the depth of the lesion, as the tissue scatters. Scattering is more present in tissue and has a larger influence on light from longer distances [34]. The resulting attenuation is considered positive for margin assessment using CLI, as it benefits superficial imaging to guide complete tumour resection.