Commissioning of a precision preclinical 200 kV x‐ray irradiator based on modular adaptations

Abstract

Background

Preclinical research in radiation oncology encompasses a range of methodologies, including in vitro cell studies and in vivo small animal experiments, as well as in silico studies to evaluate radiation‐induced side effects and tumor responses.

Purpose

This study addresses the need for high‐precision x‐ray irradiation solutions as reference for preclinical research. Modifications of an industrial kilovoltage x‐ray unit, along with the commissioning of a commercial treatment planning system (TPS), aimed to enable reliable irradiation of small animals in a horizontal beam geometry. All advancements enhancing the irradiation framework are made available, offering cost‐effective upgrades for existing systems.

Methods

An industrial kilovoltage x‐ray unit was equipped with a dual collimation system, featuring a fixed primary and variable secondary collimators with aperture diameters of 5 to 30 mm. Additional modular adaptations were designed and manufactured, including a multimodal mouse bedding unit, a dedicated dosimetry phantom and a quality assurance (QA) phantom. Output factors, percentage depth dose curves and lateral dose profiles were acquired to generate a beam model in the $\mathrm{TPS}\mu$‐RayStation 8B (RaySearch Laboratories, Stockholm, Sweden), using a diamond detector and radiochromic films. Treatment plans for 10 mice were created, evaluated via dose‐volume metrics and the homogeneity index and subsequently dosimetrically compared to QA measurements through a gamma analysis with a 1%/$1\mathrm{m}\mathrm{m}$ acceptance criterion.

Results

The resulting beam model was validated within a maximum dose deviation of 1.7%. Aperture diameters close to potential target diameters were found to be effective for achieving sufficient target coverage in silico, as demonstrated for a 5 mm target with a homogeneity index of (9.9 $\pm$ 0.7)%. Dedicated QA measurements revealed a maximum dose deviation of 1.9% from the TPS and a median gamma passing rate of 100%, confirming the suitability of the proposed solution.

Conclusions

Cost‐effective adaptations for an kilovoltage x‐ray irradiation framework were designed, manufactured and commissioned, and contribute to the accessibility of preclinical irradiation research. These components are integrated into a comprehensive preclinical particle beam platform.

1. INTRODUCTION

Preclinical research constitutes a pivotal element of radiation oncology and encompasses a diverse array of methods and techniques, ranging from in vitro experiments, involving various cell types, ¹ to animal models. ² , ³

Many radiation‐induced side effects, for example, mucositis or urinary bladder dysfunction, are investigated in animal models by irradiating parts of the body with collimated x‐ray beams. ² , ³ , ⁴ Research with respect to tumor response poses even higher requirements on the precision and conformity of the irradiation modality. ⁵ In recent years, there has been a growing interest in the field of preclinical particle therapy. ⁶ , ⁷ Preclinical studies with charged particles are motivated by a variety of open radiobiological questions, such as the prominent example of relative biological effectiveness (RBE), which describes the ratio of absorbed doses that produce the same biological effect between a reference (typically megavoltage photons) and particle irradiation. ⁸ , ⁹ , ¹⁰

Radiobiological experiments comparing multiple irradiation modalities necessitate reference irradiation in an identical experimental setting. Due to the small size and shallow position of targets in the preclinical context, the reference irradiation needs to be translated from megavoltage linear accelerators (LINACs) to dedicated kilovoltage x‐ray units. ¹¹ This growing demand of high‐precision x‐ray units triggered the evolution of dedicated kilovoltage x‐ray irradiators. ¹² , ¹³ Additional efforts to directly dock these systems to existing ion beam infrastructure have been reported. ¹⁴ , ¹⁵ , ¹⁶ Moreover, dedicated treatment planning systems (TPS) became recently available, integrating computed tomography (CT) images with a sub‐millimeter resolution. ¹⁷ Commissioning of such dedicated preclinical kilovoltage x‐ray units poses additional dosimetric challenges, as the recommendations from international guidelines cannot always be met due to deviating geometric dimensions and treatment volumes. This is of particular importance in the frame of RBE studies, as dosimetric errors in the reference field directly translate to the RBE. ¹⁸ , ¹⁹

In addition to a reference kilovoltage x‐ray irradiation unit, state‐of‐the‐art preclinical in vivo research often necessitates access to preclinical imaging modalities. Furthermore, in vivo experiments require on‐site animal housing or high‐end biosecurity measures to enable the irradiation of animals from an external animal facility. Such a fully‐equipped multidisciplinary research facility is scarce, prompting researchers to adapt and leverage the available equipment to its fullest potential.

In this study, we demonstrate the modification of an industrial kilovoltage x‐ray unit, initially not intended for preclinical research, to replicate the setup of a horizontal particle beam line for reference irradiations, including a beam collimator for various field sizes. Mimicking the same irradiation setup allows to perform preclinical experiments with identical beam incidence angles but using different beam qualities. By doing so, a more systematic comparison of the biological experiments, positioning of the underlying model and thus the resulting biological effect is given. Basic beam data for a beam model in a dedicated TPS were acquired using a down‐scaled water‐equivalent dosimetric slab phantom. For in vivo studies, a modular, low‐cost, 3D‐printed mouse bedding unit was developed and complemented by a tissue‐equivalent QA phantom for streamlined treatment verification. Finally, the entire framework was tested in a pilot treatment planning study with subsequent subject specific QA measurements.

2. MATERIAL AND METHODS

2.1. X‐ray irradiator and beam collimation

An oil‐cooled YXLON Maxishot (YXLON International GmbH, Hamburg, Germany) full‐protection kilovoltage x‐ray unit, able to deliver 10 to 200 kV beams, was used in this study. ²⁰ The x‐ray tube has been mounted to the side of the irradiation unit to mimic the experimental setting of a horizontal particle therapy beam line. The tungsten target angle was with a focal spot size of $5.5\mathrm{m}\mathrm{m}$, while filtration consisted of $3\mathrm{m}\mathrm{m}$ Be, $3\mathrm{m}\mathrm{m}$ Al, and $0.5\mathrm{m}\mathrm{m}$ Cu. The diameter of the beam exit window was $120\mathrm{m}\mathrm{m}$. All measurements were performed with $200\mathrm{k}\mathrm{V}$ and $20\mathrm{m}\mathrm{A}$—settings suitable for the irradiation of cells and small animals. ² , ⁴ The dose delivery was controlled via a timer that can be adjusted in $1\mathrm{s}$ increments.

The spectrum of the x‐ray unit was generated using the Python library SpekPy (v2.0.11.1), ²¹ , ²² , ²³ using the nominal values for tube voltage, target angle and filtration, yielding half‐value layer (HVLs) of $1.13\mathrm{mm}$(Cu) and $12.56\mathrm{mm}$(Al) (Figure A1).

For precise positioning within the unit, a vertical cross laser and a horizontal line laser (LAP GmbH Laser Applikationen, Lüneburg, Germany) have been installed inside the irradiation cabinet. To achieve variable collimation of the beam, a brass dual collimation system was implemented, comprising a primary collimator (PC) with $46\mathrm{m}\mathrm{m}$ aperture diameter and interchangeable secondary collimators (SCs) (Figure 1a). A $160\mathrm{m}\mathrm{m}$ long brass tube interconnected the PC and SC, providing a mounting mechanism for the latter. Multiple SCs with a thickness of $20\mathrm{m}\mathrm{m}$ were manufactured, covering typical aperture diameters employed in preclinical research: 5, 7, 8, 10, 15, 20, 25, and 30 mm. Computer‐aided design (CAD) files for the collimation system are available online (https://doi.org/10.5281/zenodo.12805247).

FIGURE 1.

(a) Schematic of the collimation geometry indicating the source (S), filter (F), PC, pipe (P), SC, and a phantom. The TPS reference point was defined $50\mathrm{m}\mathrm{m}$ downstream of the SC, in a depth of $10\mathrm{m}\mathrm{m}$. (b) Collimator and SFDP were mounted on the positioning stage. The phantom could be interchanged with the mouse bedding unit. (c) Schematic of the QA phantom (brown) mounted to the mouse bedding unit (grey). A detector (blue) or radiochromic films (yellow) can be inserted in the phantom for QA measurements of lateral beams. The tooth bar, which can be fixed in the nose cone to allow longitudinal fixation of the animals, is shown in green. (d) QA phantom (including inserted EBT3 films) mounted on the mouse bedding unit. PC, primary collimator; QA, quality assurance; SC, secondary collimator; SFDP, small field dosimetry phantom; TPS, treatment planning system.

2.2. Irradiation setup and QA phantoms

Modular devices were developed to enable a straightforward irradiation of small animals and dosimetry in the presented kilovoltage x‐ray unit. A high‐precision linear stage (Toplionace, Shenzhen, China) was used to align the equipment downstream of the collimator using dedicated adapter plates.

A bedding unit for mice (Figure 1c) was developed to ensure accurate and reproducible animal positioning for high‐precision in vivo studies. This bedding unit was designed to be compatible with all steps in a typical preclinical workflow, including CT, Magnetic Resonance (MR), Positron emission tomography (PET) imaging and irradiation. Key features included a nose cone and a hollow tooth bar. The tooth bar could be introduced into the nose cone to fix the animals' jaws, ensuring consistent longitudinal positioning while offering the possibility of introducing gaseous anaesthesia. The maximum diameter of the nose cone was $22.5\mathrm{m}\mathrm{m}$ and thus compatible with the small bore sizes of preclinical imaging devices.

A small field dosimetry phantom (SFDP) was designed as a down‐scaled, water‐equivalent slab phantom (Figure 1b). RW3 slabs (PTW‐Freiburg, Freiburg, Germany) with dimensions of 60 $\times$ 60 ${\mathrm{mm}}^2$ and thicknesses of 1, 5 and 10 mm, summing up to a maximum thickness of 160 mm, could be mounted within the phantom frame. Dedicated adapters for various detectors were manufactured to ensure precise alignment, as well as a consistent amount of backscattering material of 20 mm thickness, which resembles a typical preclinical scenario. Radiochromic films could be placed between the slabs at the desired depth. Lateral and longitudinal markers were included for laser alignment.

The SFDP frame and bedding unit were 3D‐printed from acrylonitrile butadiene styrene (ABS) using the fused deposition modeling (FDM) technique on a Bambu Lab X1‐Carbon (Bambu Lab EU GmbH, Frankfurt am Main, Germany) printer. ABS (Sunlu International Hong Kong Holdings Ltd., Hong Kong, China) was used due to its favorable mechanical properties and adequate dosimetric properties in x‐ray as well as proton beams. ²⁴ , ²⁵

To verify the geometric and dosimetric accuracy prior to any in vivo experiments, a QA phantom was established (Figure 1c,d). The design vaguely mimicked the shape and dimensions of a mouse, while still allowing for straightforward positioning on the bedding unit and thereby for an efficient QA workflow. The maximum dimensions were 93 $\times$ 29 $\times$ 21 mm³. The phantom featured inserts for a detector of $7\mathrm{m}\mathrm{m}$ diameter in axial orientation, as well as a stack of three radiochromic films ($30\times 30{\mathrm{mm}}^2$) in orthogonal orientation. Assuming lateral irradiation and a detector entrance window of $1\mathrm{m}\mathrm{m}$, the films and sensitive volume of the detector were located at a maximum depth of $14.4\mathrm{m}\mathrm{m}$. The entire phantom was manufactured from a solid piece of water‐equivalent material (Solid Water HE; Sun Nuclear Corporation, Melbourne, Florida, USA). CAD files for the mouse bedding unit, SFDP and QA phantom are provided online (https://doi.org/10.5281/zenodo.12805247).

2.3. Dosimetry and basic beam data

A type 30013 Farmer ionization chamber (PTW‐Freiburg, Freiburg, Germany) was calibrated at a secondary standard laboratory in an x‐ray beam at $180\mathrm{kV}$ (HVL = $1\mathrm{m}\mathrm{m}$ Cu) and $250\mathrm{k}\mathrm{V}$ (HVL = $2.6\mathrm{m}\mathrm{m}$ Cu) in terms of air kerma free in air, resulting in an identical calibration coefficient for both beam qualities. Thus, this coefficient could be used to cross‐calibrate this chamber to a PTW type 60019 microDiamond detector (mD) and a type 34045 Advanced Markus Chamber in $200\mathrm{k}\mathrm{V}$ (HVL = $1.13\mathrm{m}\mathrm{m}$ Cu) according to IAEA TRS‐398 recommendations. ¹⁹ However, not all recommendation could be fulfilled due to geometrical constraints within our device. In particular, a field size of $20\mathrm{c}\mathrm{m}$ diameter and a source‐to‐surface distance (SSD) of $331\mathrm{m}\mathrm{m}$ were used. The air kerma calibration was converted to dose to water using the mean water to air ratios of the mass energy absorption coefficients provided by Andreo et al., as well as the chamber perturbation factors provided by Bancheri et al., as outlined in the TRS‐398. ¹⁹ , ²⁶ , ²⁷

Percentage depth dose curves (PDDs) were measured using the mD, which was shown to be suitable for small kilovoltage x‐ray beams. ²⁸ For this purpose the detector was mounted in the SFDP, which was positioned $40\mathrm{mm}$ downstream of the SC, corresponding to a SSD of $331\mathrm{m}\mathrm{m}$ (Figure 1a). PDDs were acquired for all available SCs in three repetitions (without altering the setup), for depths ranging from 1 to 50 mm in steps of $5\mathrm{m}\mathrm{m}$. Output factors were measured in a depth of $20\mathrm{m}\mathrm{m}$ RW3 to replicate the reference conditions according to the IAEA TRS‐398 protocol as close as possible. Still, the preclinical setting necessitated a reference field size of $42\mathrm{m}\mathrm{m}$ for output factor measurements, as well as usage of the non‐cylindrical mD detector, deviating from the recommended configuration. ¹⁹ Dose rates for the $8\mathrm{m}\mathrm{m}$ reference aperture were measured in a depth of $10\mathrm{m}\mathrm{m}$ of RW3 at the TPS reference point (see Section 2.4) for timer settings ranging from 5 to 330 s, corresponding to accumulated doses of approximately 0.1 to 10 Gy.

To assess dose variations peripheral to the central beam axis, lateral dose profile (LDP) measurements were acquired for all apertures and six depths using Gafchromic EBT3 films (Ashland, Wayne, NJ, USA). The film batch (lot #3082203) was calibrated against the Advanced Marcus chamber for $200\mathrm{k}\mathrm{V}$ at dose levels of 0.5, 1.0, 1.5, 3.0, 5.0, 7.0, and 10.0 Gy. An Epson Expression 11000XL flatbed scanner (Seiko Epson Corporation, Nagano, Japan) was employed for image acquisition. Scanning procedures were conducted prior to and $24\mathrm{h}$ after irradiation with a resolution of $300\mathrm{dpi}$. Film handling procedures were executed following the guidelines described in the AAPM TG 235 report ²⁹ and in previous studies, ³⁰ , ³¹ achieving standard practices. Horizontal and vertical profiles were extracted from the films by averaging three individual line profiles centered symmetrically around the distributions' center, which were subsequently smoothed via a median filter (three pixels window) to reduce the influence of film noise. The resulting profiles were analyzed in terms of field size (full width at half maximum (FWHM)), penumbra, and flatness. ³² The penumbra was calculated as the distance between the points of $80\%$ and $20\%$ of maximum dose. Flatness ($F$) was calculated as the percentage difference between the maximum ($D_{\max }$) and minimum ($D_{\min }$) values across the profile within the inner 80% of the FWHM as

$$F=\frac{D_{\max }-D_{\min }}{D_{\max }+D_{\min }}\times 100.$$ (1)

2.4. Beam model

The beam model was created and validated for µ‐RayStation 8B (RaySearch Laboratories, Stockholm, Sweden), which is dedicated to the irradiation of small animals and included the following features: GPU based Monte Carlo (MC) dose engine for kilovoltage beams (v1.0, dose to water, statistical uncertainty of $0.5\%$), dose grid resolution of $0.2\mathrm{m}\mathrm{m}$, structure and aperture resolution of $0.1\mathrm{m}\mathrm{m}$. ¹⁷ Necessary beam model inputs for the TPS comprised the x‐ray spectrum (Section 2.1), as well as PDDs and relative LDPs for all available SCs (Section 2.3). Absolute dose calibration was defined via output factors relative to the $8\mathrm{m}\mathrm{m}$ aperture and a steady dose rate measured at the TPS reference point, defined $50\mathrm{m}\mathrm{m}$ downstream of the SC in a depth of $10\mathrm{m}\mathrm{m}$, at a resulting source‐to‐axis distance (SAD) of $341\mathrm{m}\mathrm{m}$ (Figure 1a).

For beam model validation, treatment plans were created on a virtual RW3 phantom for the three major apertures of 5, 8 and 30 mm diameter. All plans were scaled to a dose of $1\mathrm{Gy}$ at the TPS reference point and the predicted irradiation time served as a basis for validation measurements in depths of 1, 5, 10, 20 and 50 mm using the mD. The relative dose agreement of simulation to measurement was investigated and one‐way analysis of variance (ANOVA) was performed to identify potentially significant differences for the various depths and apertures. Test statistics were computed using the SciPy library (version 1.11.3) in Python, with significance determined for p‐values less than 0.05. ³³

2.5. Subject specific treatment planning and dosimetric verification

The translation of the dosimetric characteristics into the in vivo context was investigated for 10 BALB/cJRj mice (Ethics approval from the Federal Ministry of Education, Science and Research of Austria: 2023‐0.556.474). All animals were placed on the mouse bedding unit and received a spiral cone beam $\mu \mathrm{CT}$ scan (X‐CUBE; Molecubes, Gent, Belgium) using a $50\mathrm{k}\mathrm{V}$ tube voltage and $0.225\mathrm{m}\mathrm{A}$ current at a resolution of $0.2\times 0.2\times 0.2{\mathrm{mm}}^3$ for treatment planning. For this purpose, a Hounsfield units (HU) calibration curve was established based on tissue‐equivalent materials (CIRS 062M; Sun Nuclear Corporation, Melbourne, Florida, USA). ²⁵ Dose calculation settings were equivalent to those described in Section 2.4

A horizontal beam employing a $5\mathrm{m}\mathrm{m}$ aperture was delivered from the right side to emulate partial brain irradiation with the isocenter and dose normalization point aligned with the center of the right hippocampus. The dose prescription to the normalization point was $1.0\mathrm{Gy}$. Target coverage was assessed with three cylindrical structures (diameters of 4, 5 and 6 mm), which were centered around the normalization point and intersected with the right hemisphere of the brain. To evaluate dose in the surrounding tissue, the two brain hemispheres and the skull (threshold based 250 to maximum HU) were delineated.

Dose profiles in beam direction through the normalization point were analyzed for each treatment plan and subsequently averaged for all 10 rodents. ${\mathrm{D}}_{10\%}$ and ${\mathrm{D}}_{90\%}$ were determined additionally as surrogates for the maximum and minimum dose in the small target volumes ($<$ $0.11\mathrm{c}{\mathrm{m}}^3$). Together with ${\mathrm{D}}_{50\%}$ the homogeneity index (HI) was calculated as a metric for target coverage. ³⁴

$$HI=\frac{D_{\text{10}\%}-D_{\text{90}\%}}{D_{\text{50}\%}}$$ (2)

The generated treatment plans were re‐calculated on a mimicked $\mu \mathrm{CT}$ of the QA phantom (virtual air $\mu \mathrm{CT}$ complemented with the structures of the phantom) (Section 2.2). Each plan was subsequently delivered twice: once targeting the mD and once a stack of three EBT3 films placed within the phantom. The dose measured by the detector was compared to the prediction of the TPS in terms of relative dose deviation. The 2D dose distribution of the films was resampled to the $\mu \mathrm{CT}$ resolution and aligned to the corresponding delineated $\mu \mathrm{CT}$ structure utilizing the Python interface of the SimpleITK toolkit (version 2.3.1). ³⁵ , ³⁶ A 2D gamma analysis was then conducted to compare the doses of the TPS and film, employing the PyMedPhys library (version 0.38.0) with a $1\%$ dose difference (DD) and $1\mathrm{m}\mathrm{m}$ distance to agreement (DTA) acceptance criterion at a lower dose cutoff of $10\%$. ³⁷

3. RESULTS

3.1. X‐ray beam characteristics

The acquired PDDs, normalized to the dose measured with the reference aperture ($8\mathrm{m}\mathrm{m}$) in a depth of $20\mathrm{m}\mathrm{m}$, are shown in Figure 2a. Relative differences comparing the largest to smallest aperture increased from $17\%$ at $1\mathrm{m}\mathrm{m}$ to $37\%$ at $50\mathrm{m}\mathrm{m}$ depth. The distance from the surface to the $95\%$ dose fall‐off was $3.5\mathrm{m}\mathrm{m}$ for the $5\mathrm{m}\mathrm{m}$ aperture and $6\mathrm{m}\mathrm{m}$ for the $30\mathrm{m}\mathrm{m}$ aperture. The corresponding output factors ranged from 0.92 to 1.30 as listed in Table 1. Figure 2b shows LDPs for the reference aperture.

FIGURE 2.

(a) PDD acquired with the mD and normalized to the dose measured in $20\mathrm{m}\mathrm{m}$ depth using the reference aperture (8 mm). Each marker corresponds to three repeated measurements—the respective relative standard deviations were below $0.2\%$. (b) LDPs for the reference aperture measured with EBT3 films were normalized to the dose measured with the mD on the central axis in $20\mathrm{m}\mathrm{m}$ depth. mD, microDiamond detector; PDD, percentage depth dose curve; LDPs, lateral dose profiles.

TABLE 1.

Beam parameters in a depth of $20\mathrm{m}\mathrm{m}$ in RW3.

Aperture (mm)	Output factor	FWHM (mm)	Penumbra (mm)	Flatness (%)
5	0.92	6.1	0.6	3.6
7	0.98	8.7	0.6	4.5
8	1.00	10.0	0.7	5.0
10	1.03	12.4	0.6	4.6
15	1.12	18.6	0.8	6.0
20	1.19	24.8	0.9	6.3
25	1.25	31.4	1.2	5.9
30	1.30	37.3	1.3	7.3

Note: Output factors were normalized to the reference field employing the $8\mathrm{m}\mathrm{m}$ aperture.

Abbreviation: FWHM, full width at half maximum.

Field size, penumbra, and flatness for all apertures are shown in Table 1 for a depth of $20\mathrm{m}\mathrm{m}$. The ratio of aperture to field size was (1.24 $\pm$ 0.01), thus varying by less than 1% for all diameters. The sharpest penumbra was observed at the surface and increased with depth by up to $2\%$ and $5\%$ for the smallest and largest aperture, respectively.

Measurements for 10 timer settings of 30 to 330 s yielded a mean dose rate of ($1.722\pm 0.002)\mathrm{Gy}{\min }^{-1}$, showing a constant dose rate at a depth of $10\mathrm{m}\mathrm{m}$ for a total dose between 0.9 and 10 Gy. Increased dose rates were found for irradiation times smaller than $30\mathrm{s}$, with deviations of up to $4\%$ for $5\mathrm{s}$. Accordingly, each $1\mathrm{s}$ increment on the x‐ray unit's timer corresponds to an approximate dose increment of $0.03\mathrm{Gy}$.

A dosimetric uncertainty budget was established in accordance with the recommendations of the TRS‐398 protocol for medium energy x‐rays, ¹⁹ which is shown in Table 2. Daily output variations of the x‐ray unit were measured and accounted for with respective correction factors and are therefore excluded from the budget.

TABLE 2.

Dosimetric uncertainty budget for beam model data acquisition.

Source	Type	Value (%)
Reference chamber calibration a	B	1.4
Mass energy absorption coefficient (water to air ratio) b	B	0.5
Chamber perturbation factor c	B	1.5
Beam quality factor a	B	0.5
Detector cross‐calibration d	A	0.2
Beam data acquisition repeatability d	A	0.2
Temperature dependence of microDiamond e	B	0.1
Establishment of reference conditions f	B	1.0
Backscattering conditions (dependent of field size) g	B	0.5–4.0
Total 1‐ $\mathit{\sigma }$ standard uncertainty	Combined	2.4–4.7

Note: To account for deviating backscattering conditions during basic beam data acquisition and detector cross‐calibration, an additional uncertainty was included. The contribution of the backscattering conditions depends strongly on the field size, ranging from $0.5\%$ for field diameters $\le$ $10\mathrm{m}\mathrm{m}$ up to $4\%$ for diameters of $42\mathrm{m}\mathrm{m}$. Thus the total uncertainty is given as a range for small to large field sizes.

^a Provided by secondary standard laboratory.

^b According to TRS‐398 and Andreo et al. ¹⁹ , ²⁶

^c According to TRS‐398 and Bancheri et al., including field size dependence. ¹⁹ , ²⁷

^d Measured quantity.

^e Specified as $0.05\%{/}^{\circ }\mathrm{C}$, assuming irradiation cabinet temperature within $\pm$ .

^f According to TRS‐398, accounting for dose gradients up to $1\%/\mathrm{m}\mathrm{m}$.

^g Interpolated from tabulated values provided by Chen et al. for field diameters $\le$ $42\mathrm{m}\mathrm{m}$. ³⁸

3.2. Beam model validation

Validation for aperture diameters of 5, 8 and 30 mm at depths of 1, 5, 10, 20, and 50 mm showed relative dose deviations within −1.5% to 1.7% comparing measurements and dose calculation, indicating good agreement for all apertures and depths within the dosimetric uncertainties. A one‐way ANOVA showed no significant differences in dose agreement for the three validated apertures ($p=0.9$). However, ANOVA revealed differences in mean dose agreement for the investigated depths ($p=0.0005$), with a trend shifting from dose underestimation at $1\mathrm{m}\mathrm{m}$ to overestimation at $50\mathrm{m}\mathrm{m}$.

3.3. In silico dose calculation for mice

The treatment plans for 10 mice showed very similar dose distributions, which was reflected in a low relative standard deviation of the near minimum and maximum dose below $1.2\%$ for the target and the two brain hemispheres. The narrow confidence intervals of the averaged dose‐volume histogram (DVH) curves indicate a high anatomical similarity between the mice (Figure A2). The averaged dose profile in beam incident direction in Figure 3a reveals the highest dose depositions in the skull, with maximum values of up to $2.93\mathrm{Gy}$ for one individual (prescribed dose of $1\mathrm{Gy}$). An exemplary dose distribution in coronal view is shown in Figure 3b.

FIGURE 3.

(a) Average line profiles of dose and HU in beam direction for all mice. The confidence intervals indicate the minimum and maximum values. The peaks of the dose and HU profiles in an approximate depth of 11 and 21 mm depict the high dose deposition in the skull. Note that the horizontal axis was flipped to emphasize the beam incident direction from the right side. (b) Coronal view of a dose distribution overlayed on a $\mu \mathrm{CT}$ of an exemplary mouse. HU, Hounsfield unit.

The 95% isodose reached a depth of $5\mathrm{m}\mathrm{m}$ in the ipsilateral brain hemisphere falling to 85% in the distal part of the contralateral side. The average FWHM of the lateral profile through the isocenter was (5.97 $\pm$ 0.02) mm and was thereby in agreement with the FWHM of the commissioning measurements in RW3 as reported in section 3.1. Good target coverage was reflected by the comparable HI of (9.2 $\pm$ 0.3)% and (9.9 $\pm$ 0.7)% for the targets with 4 and 5 mm diameter. For a diameter of 6 mm the HI increased to (30.5 $\pm$ 0.8)%.

3.4. Quality assurance

Plan specific QA measurements with the mD revealed a dose underestimation of the TPS for all 10 treatment plans with a mean dose deviation of (−1.74 $\pm$ 0.04)% and a maximum absolute deviation of 1.90%. With three films per plan, a total of 30 films were irradiated, resulting in a median gamma passing rate of 100% using a passing criterion of 1%/1 mm. The lowest passing rate for an individual plan was (96 $\pm$ 3)%, while for five plans the passing rate of all three films was 100%.

4. DISCUSSION

A beam collimation system for a kilovoltage x‐ray unit with a horizontal beam was designed and manufactured and complemented by a mouse bedding unit and respective dosimetric and QA phantoms. These components are integrated as part of a preclinical platform for ion beam reference irradiation. The mouse bedding ensured reproducible animal positioning throughout the entire workflow, ranging from multimodal imaging to irradiation. The designed x‐ray collimators' primary purpose was to mimic the available particle beam irradiation geometry and by that to enable a direct comparison of the biological effect when using different beam qualities in biological experiments. ³⁹ Thus, a purely horizontal x‐ray beam line for reference irradiation was intended.

Even though the presented system is less sophisticated compared to commercial preclinical systems, we obtained comparable dosimetric characteristics. Lindsay et al., described results for a $225\mathrm{kVp}$ unit (X‐RAD 225Cx, Precision X‐Ray Inc., USA), reporting penumbrae of 0.6 to 0.7 mm for apertures of $10\mathrm{mm}$ diameter and 0.5 to 0.7 mm for aperture diameters of $5\mathrm{mm}$ in a depth of $5\mathrm{mm}$, which is in good agreement with the results presented in this study. Their study showed lower values for profile flatness, however, they also discussed the limited comparability of the flatness parameter due to its sensitivity to film processing protocols and analysis. ⁴⁰

The dose rate of our system was (1.722 $\pm$ 0.002) Gy ${\min }^{-1}$ and is thus comparable to multiple established solutions, where dose rates between 1 and 4 Gy ${\min }^{-1}$ have been reported. ¹³ , ⁴¹ , ⁴² , ⁴³ Increases in dose rate of up to $4\%$ were identified for irradiation times smaller than $30\mathrm{s}$. Consequently, this should be taken into account for dose levels below $0.9\mathrm{Gy}$.

The spectrum calculated via SpekPy was used as input for the TPS. Considering tolerances of the nominal parameters (as specified by the vendor) of the used x‐ray unit ($2\mathrm{k}\mathrm{V}$ high voltage, $0.005\mathrm{m}\mathrm{m}$ filter thickness, and target angle), we estimated an uncertainty of $2\%$ resulting in a calculated HVL of (1.13 $\pm$ 0.02) mm Cu. Respective measurements revealed a HVL of (1.09 $\pm$ 0.03) mm Cu. Although HVL calculation and measurement agree within the respective uncertainties, offsets in HVL, caused by deviations of the calculated compared to the actual spectrum, could result in inaccurate dose calculation.

Beam model validation showed good absolute agreement between dose calculation and measurement within $1.7\%$, demonstrating the suitability of the established model. Notably, dose underestimation close to the phantom surface and overestimation at depths up to $50\mathrm{m}\mathrm{m}$ was identified and was most likely caused by imperfect modeling of the x‐ray spectrum and measured beam profiles.

While the IAEA Code of Practice TRS‐398 guidelines for kilovoltage x‐ray beam dosimetry are based on the absorbed dose to water formalism, standard calibration laboratories that can provide these standards are still scarce and thus absorbed dose determination for kilovoltage x‐rays commonly still relies on air kerma calibrations, as performed in the presented study. The air kerma based measurements were converted to absorbed dose to water, as outlined in Section 2.3, this procedure introduced additional uncertainties.

The established dosimetric uncertainty budget comprises factors from the detector calibration chain, temperature dependencies of the detector response, establishment of reference conditions as well as deviating backscattering conditions and was based on the TRS‐398 recommendations. The thickness of material behind the detector to provide full backscatter is roughly equal to the field size. The in‐house developed dedicated adapter provides $20\mathrm{m}\mathrm{m}$ of backscattering material. Thus, full backscatter conditions could be assumed for the smallest field sizes $\le$ $10\mathrm{m}\mathrm{m}$, but not for field sizes $\ge$ $20\mathrm{m}\mathrm{m}$. The tabulated values of field size dependent backscatter factors provided by Chen et al. allowed us to introduce an uncertainty component accounting for this fact, resulting in an estimated maximum 1‐$\sigma$ standard uncertainty of $4.7\%$. ¹⁹ , ³⁸

Basic beam data, such as PDDs and output factors were determined using the mD. Its suitability for kilovoltage x‐ray beam dosimetry was shown by Damodar et al. However, Kaveckyte et al. also pointed out that synthetic diamond detectors show a non‐negligible variation in the relative intrinsic energy response for low energy photons. ²⁸ , ⁴⁴ The additional beam hardening filtration of the used x‐ray unit minimizes the contribution of these low energy photons to the total spectrum (Figure A1) and thus justifies usage of the mD for dosimetry in the outlined setting.

The sharp penumbra of $0.6\mathrm{m}\mathrm{m}$ for the smallest fields indicated that apertures with diameters close to potential target diameters were appropriate to achieve sufficient lateral target coverage and thus support sparing of surrounding tissues. The irradiation plans for the in silico mouse brain study were consequently generated using a $5\mathrm{m}\mathrm{m}$ SC. This beam collimation enabled good coverage of the target region, reflected in a HI of (9.9 $\pm$ 0.7)% for a 5 mm target diameter. A high anatomical similarity of the animals was reflected in the comparable dose parameters and is the ideal basis for a standardized workflow tailored to a dedicated research question.

Different aperture shapes (e.g. rectangular) might prove beneficial for healthy tissue sparing for particular anatomical sites and will be subject of future investigations. For potential target sizes below diameters of $5\mathrm{m}\mathrm{m}$, the commissioning data needs to be complemented. However, the suitability of the mD needs to be investigated for these field sizes. Alternatively, PDD and output factor measurements could be obtained with EBT3 films, thermoluminescent dosimeters or small volume scintillator detectors.

The developed QA phantom was designed to be compatible with the mouse bedding unit, allowing for a streamlined QA workflow in multiple beam modalities and eliminating the need for significant setup alterations between treatment verification and in vivo irradiation. Our QA results indicated that the outlined framework is well‐suited for in vivo experiments employing aperture diameters down to 5 mm. While the suggested phantom was aimed at practicability, more advanced, anatomical mouse phantoms have been reported previously and can be employed to additionally investigate the influence of tissue heterogeneities. ⁴⁵ , ⁴⁶

Currently, the proposed framework supports a landmark‐based positioning approach using isocentric lasers. A regular QA procedure to confirm the laser alignment with the beam isocenter has not been introduced so far. To overcome this limitation radiochromic films were used in combination with a cross‐hair collimator to validate the laser position in this study. For target diameters less than 5 mm, sub‐millimetric positioning accuracy, as provided by modern commercial solutions, is crucial. ⁴⁷ , ⁴⁸ However, achieving this accuracy will be challenging without online image‐guidance capabilities, which is the main limitation of the presented solution. Future studies will explore individual restraining jigs and potentially 3D printed body molds to address this limitation. ⁴⁹

The introduced modular adaptations can be manufactured and adapted by other users of similar systems, allowing for fast implementation and enhancements within the preclinical research community.

5. CONCLUSION

We proposed enhancements for an industrial kilovoltage x‐ray unit to enable high‐precision irradiation of small animals with a commercial TPS for dose calculation. This upgraded system includes a collimation system with interchangeable apertures, a mouse bedding unit, a dosimetric and a QA phantom. These modular designs, combined with the presented workflow and its validation, as well as the in silico study provide a cost‐effective and accessible contribution to preclinical ion beam platforms aiming for improved reproducibility, comparability, and consistency within this research field.

CONFLICT OF INTEREST STATEMENT

Dietmar Georg serves as a deputy editor for the journal Medical Physics. The Department of Radiation Oncology (Medical University of Vienna) has an institutional research contract with RaySearch Laboratories (Stockholm, Sweden).

FIGURE A1.

$200\mathrm{k}\mathrm{V}$ x‐ray spectrum calculated with SpekPy (version 2.0.11.1) using a Tungsten target with an anode angle of and $3\mathrm{m}\mathrm{m}$ Be, $3\mathrm{m}\mathrm{m}$ Al, and $0.5\mathrm{m}\mathrm{m}$ Cu filtration.

FIGURE A2.

Averaged dose‐volume histogram for 10 mice with confidence intervals indicating the minimum and maximum values.

Wolf L, Kuess P, Leitner S, Georg D, Knäusl B. Commissioning of a precision preclinical 200 kV x‐ray irradiator based on modular adaptations. Med Phys. 2025;52:4713–4722. 10.1002/mp.17758