SAPPHIRE —establishment of small animal proton and photon image-guided radiation experiments

The SmART+ IB stands on rollers in the experimental area of the UPTD and can be placed square to a horizontal beamline providing a static proton pencil beam. The correct position of the SmART+ IB is ensured every time with millimeter precision by measuring the distance of the device to the proton beam exit as well as to the wall using a digital laser measure (type GLM 50 C, Robert Bosch Power Tools GmbH, Leinfelden-Echterdingen, Germany). The device (figure 1) is equipped with a specially designed window on the back at the height of the beam pipe, allowing for integration of a proton beam. The fully shielded device features a bifocal 225 kV x-ray tube (internal filtration: 0.8 mm Beryllium) with filter slide and collimator mount, a flat panel detector (0.2 mm resolution, total area 20 × 20 cm²), a 360° rotating gantry with 30.5 cm source-axis distance, and an X–Y–Z animal stage, allowing for CBCT imaging and photon irradiation using nine different fixed nozzle collimators (circular with diameters 1 mm, 3.5 mm, 5 mm, 10 mm, 15 mm, 25 mm and square 10 × 10 mm², 20 × 20 mm², 40 × 40 mm²). A supplied 2 mm thick aluminum filter is inserted for CBCT imaging, and a 0.5 mm thick copper filter is used for photon irradiation to prevent beam hardening. For proton irradiation, a 45.82 mm thick polymethylmethacrylat (PMMA) range shifter with a water equivalent thickness (WET) of 53.16 mm for 90 MeV was utilized to be consistent with previous small animal brain studies (Suckert et al 2020, 2021b). The range shifter was attached to an aluminum collimator with a 3 mm diameter aperture placed on two linear stages (LTM80P-75-HSM, LTM80P-150-HSM, OWIS GmbH, Staufen im Breisgau, Germany) on the animal stage for lateral displacement. Monte Carlo (MC) simulations (FLUKA2020 Version 2c.6, (Fasso et al 2005, Battistoni et al 2015)) were performed to determine the optimal position of the collimator and range shifter in the beam path to minimize dose contributions by scattered protons and secondary neutrons and thus protect the shielded device, the flat panel detector and electronics (figure 2). The PRECISIOn default settings were used and the dose and fluence distributions were obtained by applying the USRBIN scorer. Based on the simulations, the collimator and range shifter were positioned as close as possible to the animal box, resulting in proton and neutron fluence as well as absorbed doses at both the flat panel detector and the SmART+ IB wall being several orders of magnitude lower than those of the central beam. The shutter of the flat panel detector efficiently protects it from scattered protons, but is transparent to neutrons. However, the secondary neutrons are emitted in a forward cone, so that the neutron exposure in the direction of the detector is expected to be relatively low. For irradiation with high doses, additional neutron shielding, preferably consisting of hydrogen-rich materials like polyethylene plates, could be placed on top of the steel shutter to minimize long-term damage to the flat panel detector. Two adapters fitting the animal stage were 3D printed from acrylonitrile butadiene styrene (ABS) for stable repositioning of phantoms used for CBCT imaging (figures 3(A1)/(A2)).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Four predefined CBCT imaging protocols with different x-ray tube voltage and current settings (40 kV with 3 mA, 60 kV with 0.5 mA and 0.9 mA, 80 kV with 0.6 mA) and a slice thickness of 0.1 mm were assessed for their image quality. These parameters are recommended by the manufacturer and were chosen to closely match those of the SAIGRT system to perform comparable animal imaging. A Hounsfield unit (HU) calibration was used to convert the CBCT image's gray values into HU values based on a linear function determined by the gray and HU values of air and water. For this purpose, the average gray values of a syringe filled with water (Ø16.4 mm) and the surrounding air were determined across all slices of the CBCT. This was performed for all CBCT imaging protocols.

For image quality evaluation, two cylindrical phantoms (head: A1, body: A2 in figure 3) made of solid water (Solid Water HE, Gammex-RMI GmbH, Gieen, Germany) with diameters of 15 mm and 30 mm, respectively, and a length of 120 mm incorporating a central 5 mm diameter drilled hole were employed. The tissue-equivalent materials for the phantoms are shown in figure 3(B) (see section 2.6). The CBCT homogeneity within these phantoms, the contrast-to-noise ratio (CNR), and the visibility of notable skull structures of a deceased laboratory mouse were examined. For all image quality evaluations, one CBCT was acquired for each imaging protocol. The homogeneity was investigated in all three spatial directions x, y, and z, inside the phantoms, which were precisely positioned at the SmART+ IB isocenter using a system of room lasers inside SmART+ IB. The CNR was determined based on the CBCT values of the same solid water phantom $C{T}_{\mathrm{water}}^{\mathrm{HU}}$ and the surrounding air $C{T}_{\mathrm{air}}^{\mathrm{HU}}$ and its standard deviation σ_air as follows:

$$\begin{eqnarray}&&CNR=\displaystyle \frac{\left|C{T}_{\mathrm{air}}^{\mathrm{HU}}-C{T}_{\mathrm{water}}^{\mathrm{HU}}\right|}{{\sigma }_{\mathrm{air}}}.\end{eqnarray} \tag{ 1 }$$

To assess the stability over time in both cylindrical phantoms, CBCT scans were acquired every half hour over a period of 2.5 h and HU values were averaged over the entire phantom. To quantify the imaging dose per mouse for the imaging protocol with the best image quality according to the selected parameters, the soft x-ray chamber (model 23342, PTW Freiburg, Germany) was positioned on the animal stage at isocenter position. The soft x-ray chamber was irradiated with the settings from the best imaging protocol once with the x-ray tube in the upper position and once in the lower position through the animal stage, and the dose rate and ramp-up effect were measured to derive the imaging dose under rotation of the gantry. To examine the impact of the animal box on CT numbers, all tissue-equivalent materials were positioned within the head phantom, itself situated inside the animal box. Subsequently, a CBCT was conducted for each material using the best imaging protocol.

To characterize the photon beam, depth dose profiles and lateral profiles were acquired for seven collimators at the SmART+ IB's isocenter operating at 200 kV, 20 mA, with the 0.5 mm copper filter (half-value layer (HVL): 1.03 mm Cu). Figure 3(C) shows the setup for the beam measurements, in which GafChromic EBT3 films (Ashland Inc., Wilmington, Delaware, USA) were placed inside a solid water slab phantom (type 451, Gammex-RMI GmbH, Gieen, Germany) at different depths for both 20 s and 80 s exposures.

The EBT3 films were calibrated in homogeneous radiation fields of 200 kV x-rays against a Farmer ionization chamber (type 30010, PTW Freiburg, Germany) and spread-out Bragg-peak protons at UPTD (Beyreuther et al 2019). Two days after irradiation, the films were scanned with a flatbed scanner (model Expression 11000XL, Seiko Epson Corporation, Düsseldorf, Germany) at 300 dpi and analyzed by a Python script (Python Software Foundation, www.anaconda.org, version 2.7) to calculate the absorbed dose to water D_w from the net optical density netOD:

$$\begin{eqnarray}&&{D}_{{\rm{w}}}(netOD)=\displaystyle \frac{{p}_{1}\cdot {netO}{D}^{2}+{p}_{2}\cdot {netOD}+{p}_{3}}{{netO}{D}^{2}+{q}_{1}\cdot {netOD}+{q}_{2}},\end{eqnarray} \tag{ 2 }$$

with p₁, p₂, p₃, q₁ and q₂ incorporating fit parameters.

An essential component of the SAPPHIRE project for conducting preclinical in vivo experiments with small animals is a comprehensive QA. To ensure a reproducible and precise alignment of the proton beam, collimator, and target, a device-specific QA protocol was developed, which must be followed before starting the experiments including several QA phantoms.

To characterize the evolution of the proton beam in air and its dose distribution in water, all SAPPHIRE components necessary for proton irradiation were assembled. Lateral and depth dose profiles were acquired at four different positions along the proton beam trajectory using an energy of 90 MeV. These positions encompassed the following locations: immediately behind the beam exit (at a distance of 5 cm), both preceding and following the range shifter, and directly behind the collimator. For the measurement of lateral and depth dose profiles in a water-equivalent target, a film stack consisting of 45 films (WET of 16.2 mm) was irradiated at each position. Film evaluation was performed in the same manner as described in equation (2).

Additional measurements of the lateral profiles were employed with a microDiamond detector (model 60019, PTW Freiburg, Germany) as described in Horst et al (2023), while for the laterally integrated depth dose profiles, the Giraffe multi-layer ionization chamber (IBA Dosimetry, Schwarzenbruck, Germany) was utilized with the same beam parameters.

For a precise prediction of proton ranges on basis of CBCT images of SmART+ IB, a device and system-specific HLUT needs to be created defining the relationship between HU values and proton stopping power ratio (SPR) relative to water. To create this HLUT, 19 different cylindrical tissue-equivalent inserts (5 mm diameter, 30 mm length) were placed inside the head and body phantoms (see figures 3(A)/(B)). The assignment of the materials and their material properties can be found in the supplementary materials in table S1. For each phantom, a CBCT was taken with the best imaging protocol, and the SPRs were extracted from the work of Peters (2022). This empirical method for determining the HLUT was subsequently validated by the stoichiometric method of Schneider et al (1996) including 71 human tissues as well as the 19 materials. The influence of the animal box on the HLUT due to the shifted CT numbers was also assessed.

The linear function parameters for the current HU calibration for all four imaging protocols are outlined in the supplementary materials in table S2, enabling the conversion of the CBCT's grey values into HU values. The cupping effect (Brooks et al 1976) in the x-direction of the SmART+ IB is shown in figure 4(A) for both head and body phantoms across four imaging protocols. Each plot includes a quadratic fit as a trendline overlay. Deviations in HU values between the edge and center of the phantoms are 15 HU for the head phantom and 40 HU for the body phantom, well below the CT number standard deviations of the inserts. This corresponds to a proton range uncertainty of <0.1 mm (<0.3 mm) for the head (body) phantom. These range uncertainties were calculated using the empirical HLUT of the 60 kV, 0.5 mA imaging protocol, in which deviations in HU values for solid water correspond to discrepancies in SPR values, hence resulting in range uncertainties. The body phantom profiles exhibit similar discrepancies to those observed in the head phantom. Many curves show a noticeable valley around the center, indicating a CBCT reconstruction artifact. The profiles in the y-direction have a similar pattern to the depicted x-profiles, while in the z-direction, there is no observed cupping effect (see supplementary materials figure S1/S2).

Zoom In Zoom Out Reset image size

To minimize the imaging dose while upholding high image quality, the 60 kV, 0.5 mA (HVL: 1.85 mm Al) imaging protocol, with an imaging dose of less than 26 mGy, is set as the standard for future mouse experiments. A corresponding sagittal view of the midsection of a deceased laboratory mouse skull is displayed in figure 4(B). With acceptable low HU value deviations due to the cupping effect and a CNR of 25.48 ± 2.38 for the head phantom and 22.18 ± 2.20 for the body phantom, which allows for clear differentiation of all critical structures, the protocol proves to be suitable.

In figure 4(C), the CT number differences for individual CT number measurements in water and air are depicted relative to their first measurements at time t = 0 min for both the head and body phantoms over a period of 2.5 h. It can be observed that the CT number difference gradually increases over time due to the afterglow of the flat panel detector; however, this does not have any significant impact on the range prediction. The influence of the animal box on the CT numbers of the tissue equivalent materials was investigated for the 60 kV, 0.5 mA imaging protocol. The differences in CT numbers with and without the animal box can be observed in figure 4(D), indicating a linear relationship.

In figure 5, both cross-plane (x-profile) and in-plane (y-profile) profiles are shown for seven different collimators in 10 mm (water) depth (figures 5(A)) and at various depths along the beam path for the Ø3.5 mm collimator (figure 5(B)). The dose values were consistently averaged (median) over five EBT3 films for each time and depth of the solid water slab phantom from figure 3(C). The corresponding full width at half maximum (FWHM) and penumbra values can be found in table 1. Gaussian fitting was applied for the Ø1 mm collimator (R² = 0.992), while sigmoid edge fitting was used for the remaining collimators (R² > 0.990). This involved utilizing 50% of the central beam dose for FWHM and 20%/80% of the central beam dose for the penumbra calculations. All collimators exhibit expected FWHMs and a slight increase in penumbra with larger collimators. For the Ø3.5 mm collimator, both the FWHM and penumbra increase linearly with depth for both profiles, as anticipated.

Zoom In Zoom Out Reset image size

The dose rates for all investigated collimators can be found in table 1. The dose rate was calculated from the difference between the 80 s and 20 s exposed films, each including all dose values from a 9 × 9 px² (0.76 × 0.76 mm²) field centered around the center of mass. It is evident that for larger collimators, the available dose rate is higher. Based on the dose rate values, relative depth dose profiles were determined and illustrated in figure 6, with normalization to a depth of 10 mm due to the greater fluctuation of surface dose values.

Zoom In Zoom Out Reset image size

A QA protocol (figure 7(A)) to be followed before each experiment is implemented for successful alignment of the collimated proton beam and the SmART+ IB isocenter using a custom-made QA phantom (50 mm × 50 mm × 87 mm) constructed from PMMA, as illustrated in figures 7(B)/(C).

Zoom In Zoom Out Reset image size

For the alignment of proton beam and collimator, the SmART+ IB is positioned in front of the proton beamline with the small proton beam window facing the beamline (figure 1). The QA phantom for collimator alignment is placed on to the animal stage as shown in figure 7(B). This phantom features a specially designed edge at the bottom for secure placement on the adapter. At the center of the phantom's surface facing the beam, a small steel ball with a 2 mm diameter is embedded to serve as the target for the QA procedure. EBT3 films can be placed on the same side of the phantom as the steel ball and within a 5° inclined cut trough the phantom for lateral beam profile and depth dose profile measurements, inspired from the work of Kim et al (2019). It must be ensured that the QA phantom is targeted by the proton beam when it is positioned at the SmART+ IB isocenter by using the integrated room laser system. A new EBT3 film is positioned on the front side of the phantom (note that the collimator holding the range shifter is not yet placed in the beam path). Afterwards, the proton beam window and the opposite door are closed, and a CBCT of the QA phantom is acquired using the 60 kV, 0.5 mA imaging protocol. The animal stage coordinates are ${P}_{1}({x}_{{P}_{1}}={y}_{{P}_{1}}={z}_{{P}_{1}}=0)$. The center of the steel ball (target) is identified in the CBCT, and the animal stage is then moved to position the steel ball in the isocenter resulting in new stage coordinates ${P}_{2}({x}_{{P}_{2}},{y}_{{P}_{2}},{z}_{{P}_{2}})$. Next, with the proton beam window and door opened again, a proton beam is delivered using 90 MeV, 0.1 nA for 17 s to produce an image of the beam spot (figure 7(D), black dashed circle). The collimator, along with the PMMA range shifter, is then introduced into the beam path, directly in front of the QA phantom with collimator axis coordinates ${Q}_{1}({y}_{{Q}_{1}},{z}_{{Q}_{1}})$. A second proton beam is delivered using 90 MeV, 0.1 nA for 80 s (figure 7(D), red dashed circle). The position of the steel ball is marked with a pen on the film (figure 7(D), white dashed circle), and the film is scanned using the flatbed scanner. The shift S₁ between the center of the two dark patches from the proton beam is measured using Fiji ImageJ (version v1.53n, 64-bit Windows, Schindelin et al (2012)). Based on this, the collimator linear stages are adjusted, and the new coordinates ${Q}_{2}({y}_{{Q}_{2}},{z}_{{Q}_{2}})$ are recorded, ensuring that the collimator is now aligned with the proton beam. To verify the alignment of the proton beam and collimator, the procedure is repeated. The shift S₂ between the proton beam spot without the collimator and the marked (isocenter) position of the QA phantom's steel ball is also determined from the EBT3 film.

In order to align the collimated proton beam with the target for irradiation (steel ball), the rectangular phantom is used, as shown in figure 7(E). The steel ball of the rectangular phantom is located between two plastic plates, with an EBT3 film placed directly behind it. With the rectangular phantom in the animal box, a CBCT is acquired using the same imaging parameters as before. The center of the steel ball within the CBCT is used as target position and moved to the isocenter, resulting in new stage coordinates ${P}_{3}({x}_{{P}_{2}},{y}_{{P}_{3}},{z}_{{P}_{3}})$. Then, the animal stage is shifted by S₂, and the collimator stages are shifted by −S₂ and the difference between P₂ and P₃. With accurate execution, the proton beam, collimator, and steel ball are now properly aligned. The result in figure 7(F) is obtained after delivering another proton beam using 90 MeV, 0.1 nA for 50 s, yielding a black patch (indicating dose on the EBT3 film) with an unexposed spot in the center (representing the dose spared by the ball). The deviation of the target from the actually irradiated spot is in this case <0.35 mm.

For multi-day mouse experiments, the SAPPHIRE alignment procedure of proton beam and collimator takes approximately 45 min and needs to be performed once per beam time, whereas the verification with the rectangular phantom should be conducted every morning prior to the initial animal irradiation. The targeting steps for the irradiation of mice will be identical to those of the rectangular phantom, targeting an anatomical structure instead of the steel ball.

Figure 8 presents the relative lateral dose profiles in both the horizontal x- and vertical y-direction of the proton beam behind the Ø3 mm collimator at various depths in water. These profiles were measured using both EBT3 films (relative measurement uncertainties <6%) and the microDiamond detector (relative measurement uncertainties <1.5%) and show good agreement. Corresponding FWHM and penumbra values for the film measurements can be found in table 2. As expected, both FWHM and penumbra increase with depth. Sigmoid edge fitting was applied (R² > 0.989), utilizing 50% of the central beam dose for FWHM and 20%/80% of the central beam dose for the penumbra calculations. Both horizontal directions (x-profile) and vertical directions (y-profile) exhibit similar FWHM trends, with the vertical direction showing a slightly sharper penumbra.

Zoom In Zoom Out Reset image size

The laterally integrated depth dose profiles obtained from the Giraffe (relative measurement uncertainties <0.1%), EBT3 films, and FLUKA simulations are illustrated in figure 9. FLUKA parameters (range shifter thickness and energy spread) were fine-tuned, and simulations were executed iteratively to match the Giraffe measurements, as described in Parodi et al (2012) for clinical proton beams. The energy spread obtained from this iterative optimization (flat energy distribution with a width of 1.28% centered at 90 MeV) fits well to the characteristics of the energy selection system in our proton therapy system. With the fine-tuned FLUKA model, a depth profile of the dose-averaged linear energy transfer (LET_d) in water was calculated. Utilizing the LET_d profile, corrections were applied to the EBT3 film measurements (see supplementary materials figure S3) due to the under-response U of EBT3 films near the Bragg peak (Anderson et al 2019) as follows:

$$\begin{eqnarray}&&U=-0.0251\,\mu {\rm{m}}\,{\mathrm{keV}}^{-1}\cdot LE{T}_{{\rm{d}}}+1.02.\end{eqnarray} \tag{ 3 }$$

It is worth noting that the function for under-response is based only on LET_d values up to 9.27 keV μm⁻¹, while values up to 17 keV μm⁻¹ occurred in the measurements in this work. For these higher LET_d regions, equation (3) was simply linearly extrapolated. The dose values from the Giraffe measurement are underrepresented in the Bragg peak compared to the EBT3 film measurements, corroborating the findings of Bäumer et al (2015). The dose rate at the Bragg peak (beam current 0.06 nA), calculated as the mean of all LET_d-corrected EBT3 film dose values within a 9 × 9 px² field centered around the center of mass, is 17.72 ± 0.35 Gy min^-1. The distal 80% depth is 12.72 mm, and the 80%–20% fall-off distance is 1.13 mm, both obtained through the Bortfeld fit (Bortfeld et al 1996) of the Giraffe measurement (R² = 0.993). While there is a range shift between the uncorrected depth dose profile of the EBT3 films and the Giraffe profile, it almost disappears when the LET_d correction is taken into account.

Zoom In Zoom Out Reset image size

The HLUT for the 60 kV, 0.5 mA imaging protocol for the head and body phantom is shown in figure 10, created using both the empirical and stoichiometric method. A similar trend is observed for the soft tissues throughout both methods and phantoms, but for the low and high CT number tissues, the methods diverge. Also noticeable is the nearly identical trend of the individual calibration curves between the phantoms. The effect of the animal box on the HLUT for both phantoms is similary evident, as it rises concerning the increasing CT numbers in relation to the empirical method. Ultimately, the HLUT for the treatment planning and proton dose calcualtion of mouse experiments is not yet determined and will be specified for each experiment with the help of μRayStation. As an example, the empirical method has been validated for the 60 kV, 0.5 mA imaging protocol in the head phantom with three unknown materials, achieving an uncertainty of the distal 80% depth of Δr₈₀ < 0.51 mm (4% of the residual range of the proton beam behind the range shifter and collimator).

Zoom In Zoom Out Reset image size