Molecular imaging using the theranostic agent ^197(m)Hg: phantom measurements and Monte Carlo simulations

The physical fundament of nuclear medicine is the application of radioactive isotopes. The emitted radiation can be used either for diagnostic or for therapeutic purposes. For diagnostics, it is favorable that the isotopes emit gamma radiation with energies of a few hundred kiloelectron volt like ^99mTc, ¹²³I, or ¹¹¹In. For therapeutic application, alpha- or beta-particle emitting radionuclides are applied to achieve a large dose deposition in tumors or metastases like ²²³Ra, ⁹⁰Y, ¹⁷⁷Lu, and ¹⁸⁸Re.

Prior to therapeutic treatment, it is mandatory to estimate the absorbed dose. To this end, pre-therapeutic imaging is performed in order to visualize the distribution of lesions and to estimate the biokinetics of the radionuclides. It is favorable that the radionuclides for diagnostic and therapeutic application are isotopes of the same chemical element because little variations in the chemical composition of radiotracers can cause large variations in the biodistribution during diagnostic and therapeutic application. Hence, a special focus is on the development of “theranostic” radioisotopes that combine the advantages of therapeutic and diagnostic radioisotopes.

One such radioisotope is mercury ¹⁹⁷Hg and its metastable state ^197mHg, henceforth referred to as ^197(m)Hg. Radiomercury in the form of ²⁰³Hg was initially investigated for clinical use in the 1950s [1, 2] and a few diagnostic applications of ¹⁹⁷Hg occurred in the 1970s [3, 4]. Back then, radiomercury was produced using reactor neutrons and had a low specific activity. Consequently, a large amount of mercury was required for imaging, leading to neurotoxic side effects [5, 6]. The novel possibility of producing ^197(m)Hg using a cyclotron allows to achieve high specific activities without any neuro- or chemotoxic effects [7]. Hence, ^197(m)Hg is very attractive for therapeutic and diagnostic application.

Therapeutic applications of ^197mHg and ¹⁹⁷Hg are conceivable because of their suitable emission spectra [8]. The radial dose distribution is comparable to the dose distribution from ¹⁷⁷Lu [7], a therapeutic radionuclide that is routinely used in nuclear medicine [9]. Apart from its therapeutic potential, ^197mHg is suitable for imaging studies due to gamma and X-ray emission upon its decay. This could, for example, allow measurements of its biodistribution, a prerequisite for dose calculations in the therapeutic setting.

This manuscript focuses on the imaging capabilities of ^197(m)Hg. To evaluate the potential for diagnostic imaging, phantom measurements were performed and compared to Monte Carlo simulations in order to better understand the physical behavior of the photons emitted by ^197(m)Hg. This is the first step for future research and the transfer to in vitro and in vivo application of ^197(m)Hg-labeled radiotracers.

In each of the following subsections, we will compare the results of measurements and simulations. To clarify our meaning, we stated if we address measurement or simulation.

Energy spectra

Figure 4a contrasts the measured and simulated spectra for this sample. The simulated spectrum was supplemented with a measured background distribution whose overall normalization was chosen so as to match the counts in the high-energy sideband (above 320 keV) of the measured spectrum. The simulated contributions of the two isomers were first normalized relative to one another using R _a and then normalized together such that the overall simulated counts match those of the measured spectrum with the normalized background removed.

Relative FWHM energy resolutions measured from the 134- and 279-keV photo-peaks of a ^197(m)Hg source with R _a  = 2.1 were 9.6 and 9.1%, respectively (Fig. 4b). Together with the 365- and 637-keV signals from a measured ¹³¹I sample, an energy dependence was observed which was linear to a very good approximation (Fig. 4c). The measured resolution compares well to the value stated by the vendor as ≤ 9.6% at 140 keV [13].

Photon scattering in different compartments was obtained by separately tallying the different events in the simulation. Using this information, the simulated spectrum can be decomposed into photon origins. This is done for a spectrum in the tomographic setup in Fig. 5 in which the energy windows are also indicated. The window at 70 keV contains significant amounts of scattered photons, as does the one at 135 keV. Their sources are different, however. While the former window is mostly scattered into from the phantom or table, the scatter contribution in the latter primarily originates from behind the NaI crystal. The window at 280 keV is largely free of scattered events.

Planar imaging

A comparison of measured and simulated images acquired with the bar phantom and both collimators can be seen in Fig. 6. Both collimators are capable of resolving the lowest spatial frequency segments of the bar pattern, while the highest frequency parts are not resolved. This enables one to find crude resolution limits by inspection. They are located at 6 and 10 mm for the LEHR and HEGP collimators, respectively, at small source-to-collimator distances. The figure also shows the raw simulation results in the rightmost column, which do not yet incorporate the intrinsic resolution model. Even the finest modulation pattern can be clearly resolved with the LEHR collimator, while imaging with the HEGP collimator mostly resolves its hexagonal hole pattern.

The procedure for determining the intrinsic MTF is now employed using measured and simulated LEHR dataset recorded in the energy window centered on 70 keV. First, the individual modulation patterns are projected onto a line running along the direction of modulation yielding profiles through the modulation patterns. Minimal and maximal amplitudes were taken from Gaussian fits to the profiles’ peaks and troughs to smooth out statistical fluctuations. This was repeated for the four modulation patterns and the intrinsic MTF calculated via Eq. 1. The approximated intrinsic MTF was parameterized by a width parameter ω = 0.0873 mm^− 1, which corresponds to an FWHM in the spatial domain of 4.29 mm (σ = 1.82 mm). The results of applying this spatial resolution model on the raw simulated data can be seen in the middle column of Fig. 6, which now shows an encouraging correspondence to the measured images.

A comparison of these profiles between measurement and simulation is presented in Fig. 7. The reproduction of the resolvable patterns is well modeled for either collimator, and one can more clearly identify the Moiré pattern exhibited by the 6-mm quadrant imaged with the HEGP collimator.

Figure 8 shows the contributions to the final LEHR image from each of the energy windows. One can observe progressive deterioration of image quality with increasing energy both in measurement and in simulation.

In the window at 135 keV, this is due to backscattered photons as demonstrated in the previous section, while image broadening in the 280-keV window could be a consequence of septal penetration. The relative contributions from these two windows to the combined image were 10 and 20%, respectively, while the window centered at 70 keV accounts for the bulk of the image. This composition depends on the time passed since production of ^197(m)Hg, and earlier measurements would have had larger contributions from the higher energy windows. Using simulated data, the issue of septal penetration can be investigated further. Figure 9 shows the septal penetration rate for three source energies that was obtained from the simulation. While the penetration behavior of the 70- and 134-keV samples is very similar, the high-energy component has a tendency to penetrate more septa, as is to be expected. There is only a slight shift towards more penetration for the HEGP collimator, while the number of penetrated septa has its most probable value at 10 in the LEHR case.

Tomographic imaging

Figure 10 shows the geometric mean of the anterior and posterior projections of the tomographic setup for measurement and simulations. The positions and relative intensities of the three sources agree quite well, but the simulated image shows a fair amount of statistical noise. While the neck structure of the vial sources is hinted at in the measured projection, it cannot be made out in the simulated one. The figure also shows the reconstructed central axial slice, and again, positions and intensities agree quite well overall. To investigate the spatial resolution of the reconstruction, profiles connecting the centroid source positions were extracted. The inter-source distances agree to within a few millimeters, while the inner diameters of the vial sources reconstructed from SPECT measurements are within 5%, and the simulated ones within 10%, of the physically measured values. In both cases, a slight underestimation occurs. The SYR source, on the other hand, is reconstructed with a width more than twice the actual value, indicating that the resolution limit is somewhere between the SYR and V11 diameters, i.e., between 5 and 19 mm.

The first application of radioactive mercury isotopes for nuclear medicine imaging was 60 years ago [1, 2]. However, radiomercury became unappealing due to neurotoxic side effects [5, 6]. These side effects were caused by the large amount of substance that was required for imaging due to the low specific activity of reactor produced radiomercury. Cyclotron-produced ^197(m)Hg allows now imaging with high specific activity as wells as the availability of a large amount of the isomeric state ^197mHg that is very prospecting for therapeutic application [7]. ^197mHg has a half-life of 23.8 h and emits monoenergetic conversion electrons with energies of 150 keV (50%) and 119 keV (33%). These electrons have a maximal range in water of 0.3 mm [7, 16]. Hence, prospectively, a steep dose gradient is expected in tumors containing ^197mHg. Previous dose calculations revealed that the dose per decay from ^197(m)Hg is 1.89 times greater than the dose from ¹⁷⁷Lu in a spherical lesion with a mass of 10 g [17]. In addition, ^197mHg and ¹⁹⁷Hg emit low-energy Auger electrons with a maximum range of about 1 μm that makes ^197(m)Hg attractive to radiobiological cell experiments. Whereas future therapeutic applications are promising, in the first step, the imaging capabilities of ^197(m)Hg were studied to evaluate if this isotope is valuable for molecular imaging or theranostic applications.

The scope of our work was to investigate the imaging capabilities of radiomercury ^197(m)Hg under experimental conditions and to compare the results with Monte Carlo simulations. Due to the complex emission spectrum of ^197(m)Hg with simultaneous emission of gammas and X-rays in the range of 50–300 keV, Monte Carlo simulations offer an option to explore the physical processes that determine imaging performance. Therefore, we used the software GATE that was previously used by other groups to simulate the imaging processes for SPECT and PET systems [11, 18–20].

Energy resolution of the Philips BrightView gamma camera was measured at four energies and found to not scale with the frequently assumed \( 1/\sqrt{E} \)energy dependence. This behavior is caused by other sources of variation in output pulse amplitude beside statistical fluctuations, for example, nonuniform sensitivity of NaI crystal or nonuniform light collection efficiency [15]. A linear relationship was adopted instead, and the comparison of measured and simulated spectra (Fig. 4a) shows a good recreation of widths and positions of photo-peaks. There is also good agreement in the intermediate energy regions, mostly populated by background events and photons, which were inelastically scattered. The correct incorporation of this latter signal provides assurance that the material geometries have been modeled sufficiently accurately. The good representation of the mid-energy peak and the contribution of backscattered high-energy photons as shown in simulations (Fig. 5), in particular, demonstrate that the BSC was adequately described, despite a lack of detailed knowledge regarding its exact geometry. Concerning photo-peak amplitudes, one can see a good general agreement in Fig. 4a. While the lack of an absolute normalization makes it hard to ascertain which energy region is best represented in terms of intensity, the current scheme hints at a slight overrepresentation of the high-energy emissions, near-perfect agreement in the mid-energy domain, and reasonable correspondence for the low-energy cluster. Some discrepancies do, however, become apparent in the latter. While there is a marked shoulder structure at its high end (75–80 keV) in the measurement, this behavior is only hinted at in simulations. This region of the spectrum is populated by a Gamma emission from the ground state (77 keV), as well as K _β X-ray emissions from both isomers (average energies 80 and 78 keV). All of these were represented with 77 keV photons in the simulations, resulting in a lack of detail. Considering the energy window at 70 keV as a whole; however, the accuracy was judged sufficient to evaluate imaging capabilities. Lead X-rays from the collimator can interfere with gamma-ray emission in that energy window and were modeled in the simulation.

A realistically modeled spatial resolution is likely to be of greater importance for imaging studies. Extraction of the intrinsic resolution of the scintillator-PMT system via the method chosen here is limited by the range of energies, modulation directions, and positions considered. Detailed measurements of the pure intrinsic component with a uniformly illuminated multi-slit plate at different energies and orientations would have been preferable, but even in their absence, correspondence seems sufficient. Comparing modulation patterns in the plane setup (Fig. 7) revealed a very good agreement between measurement and simulation as well with the vendor specification. This comparison is made at minimal source-to-collimator distances, where the intrinsic component to system resolution is significant. At farther distances, collimator resolution becomes the limiting component. Resolution also compares reasonably well in this domain (Fig. 10), albeit including the obfuscating influence of scattered events in the tomographic setup. Downscatter corrections are important especially when using the low-energy window. Hence, for correct SPECT quantification of ^197(m)Hg, attenuation and scatter correction models need to be implemented by camera manufacturers.

As far as ^197(m)Hg imaging is concerned, one can see that it is largely dominated by emissions in the low-energy range around 70 keV. This is partly due to the higher emission yield in this region, which increases further as the sample ages. Another aspect favoring the low-energy range is its compatibility with LEHR collimators. Since they are usually designed for ^99mTc imaging, one would expect the mid-energy region near 134 keV to also be suitable for imaging, but this region contains a non-negligible contribution of backscattered high-energy photons. Low-energy emissions do suffer from greater attenuation, however, so it remains to be seen whether they can be usefully exploited in a realistic clinical scenario.

Our study has two limitations. On the one hand, no calibrated measurement device was available to direct measure the activity of ¹⁹⁷Hg and ^197mHg. Hence, we could only estimate the activity by using the calibration factor of ¹²³I for our dose calibrator. Nevertheless, we considered the activity ratio R _a that was measured precisely by gamma spectroscopy. On the other hand, no medium energy collimator was available for phantom measurements, although it could be expected that this collimator would be most suitable for imaging. Despite those methodical limitations, the main results were not affected.

^197mHg is attractive for therapeutic application due to the emission of low-energy Auger and conversion electrons. Imaging is possible with both radionuclides but especially with ¹⁹⁷Hg. Hence, pure diagnostic tracers could be labeled with ¹⁹⁷Hg alone to reduce the radiotoxicity for patients. This is an advantage in comparison to the beta-particle emitting radionuclide ¹⁷⁷Lu that is used for therapy with only 17% gamma emission probability.

In conclusion, we demonstrated well the feasibility of ^197(m)Hg for molecular imaging. This was investigated by measurements and verified by Monte Carlo simulations. Thereby, the suitability of GATE for Monte Carlo simulation of a Philips BrightView camera was demonstrated. Molecular imaging of ^197(m)Hg is a prerequisite for in vivo dosimetry during future application in the treatment of micrometastases and small tumors.