ABSTRACT. Breast cancer remains one of the most prevalent cancers affecting women worldwide, and radiotherapy plays a crucial role in its clinical management – as adjuvant therapy following surgery or as a primary treatment modality for inoperable cases. Dose fractionation is a central aspect of radiotherapy with substantial implications for treatment efficacy and safety. Whereas conventional fractionation schedules with daily doses of 1.8-2 Gy constituted the clinical standard over decades, the advent of improved treatment planning algorithms and high precision dose administration workflows enabled the implementation of hypofractionated regimens with fraction sizes of > 5 Gy, revealing non-inferior efficacy and safety outcomes in clinical trials. So far, the selection of the fractionation regimen does not take into account the radiobiological characteristics of the respective tumor. This leaves the optimal, personalized fractionation regimen for patients potentially unrecognized. Therefore, we are investigating the influence of dose fractionation in clinically relevant regimens on the clonogenic survival of various breast cancer subtypes in vitro and in orthotopic mouse xenotransplants. We made use of a panel of breast cancer cell lines and measured clonogenic survival upon irradiation in both single-shot and different fractionation regimens. To assess fractionation-dependent recovery, dose equivalents of single-shot irradiation were calculated, and individual recovery scores were extracted via dimensionality reduction. These scores were integrated with exome and transcriptome sequencing data and time course profiles of radiotherapy-induced cell fate decisions. Significant differences in overall radiosensitivity and fractionation-dependent recovery were observed. Hierarchical clustering based on the dose equivalents of the fractionation regimens revealed clusters of distinct fractionation-dependent recovery which were associated with the genetic status of DNA damage repair regulators, the transcriptomic breast cancer subtype, and the pattern of radiation-induced cell fate decisions. Representative cell lines of the recovery clusters were orthotopically transplanted into athymic mice and subjected to single-dose or fractionated radiotherapy, respectively. Evaluation of the therapeutic outcome according to RECIST1.1 criteria basically confirmed the in vitro observations. Our work reveals an unexpectedly high degree of heterogeneity in fractionation-dependent recovery in preclinical models of breast cancer and characterizes underlying molecular and cell biological determinants. These findings open the perspective of personalized fractionation schedules in the future and identify vulnerabilities for pharmacological modulation of fractionation-dependent recovery which deserve further investigation.

ABSTRACT. Introduction Off-target cardiac irradiation during cancer radiotherapy can invoke chronic, potentially lethal cardiotoxicity, but current diagnostics focusing on contractile dysfunction manifest too late for effective intervention. We are therefore evaluating molecular imaging agents developed in-house which target microvascular damage, fibrosis, and mitochondrial dysfunction to detect toxicity earlier and help develop cardioprotective strategies[1-4]. To enable this, we are first establishing a clinically relevant rodent model of radiation-induced cardiotoxicity.

Methods 28 male Wistar rats (11 weeks, with age-matched CT-only controls) received 18 Gy whole heart CT-guided irradiation (SmART+, Precision X-ray Irradiation). Animals were irradiated with two sets of parallel-opposed 10 mm x 10 mm X-ray beams (225 kVp, 20 mA, 0.3 mm Cu-filtered) entering laterally to the rat. Of several treatment plans tested, this beam arrangement provided the best compromise between heart coverage while minimising dose to lung and spinal cord. Longitudinal echo- and electrocardiography (Vevo F2, Visual Sonics) and CT imaging were used to monitor cardiac structure and function over time. At periodic milestones, animals were culled to assess microvascular damage, fibrosis, and inflammation by histology of mid-ventricular heart and coronal lung slices. Biomarkers of hypoxia, mitochondrial function, senescence, and inflammation will be assayed in apical tissue alongside metabolomic profiling by NMR spectroscopy. Finally, longitudinal blood samples will be assayed for a panel of cytokines and cardiac troponin, a conventional marker of cardiac injury.

Results/Discussion This is an ongoing longitudinal study of chronic injury. Echocardiography showed progression of injury over 20 weeks with a peak increase in left ventricular anterior wall (LVAW) thickness and fractional shortening (FS) with reduced end systolic/diastolic (ESV/EDV) volumes at 10 weeks post-irradiation. This suggests myocardial thickening and fibrotic remodelling and increased contractility in response to injury. After 10 weeks, the reverse occurs, likely due to atrophy. To date, one irradiated rat exhibited arrhythmia suggestive of atrioventricular block. Our longitudinal ECG monitoring will determine if conduction abnormalities are a feature of this model, as they are clinically. We are currently performing analysis of global/regional cardiac strain, which has emerged clinically as a robust and earlier indicator of evolving cardiotoxicity.

Conclusion Our model of radiation-induced cardiotoxicity exhibits relevant clinical patterns of functional myocardial remodelling in response to a single fraction 18 Gy whole heart irradiation. In-depth characterisation of molecular responses in tissues will be compared with functional evolution to test correlative relationships and inform upcoming imaging studies and potential cardioprotective strategies.

References [1] Medina, R.A., et al. ‘64Cu-CTS: A Promising Radiopharmaceutical for the Identification of Low-Grade Cardiac Hypoxia by PET’, J Nucl Med, 2015/56, 921-926. [2] Safee, Z.M., et al. ‘Detection of anthracycline-induced cardiotoxicity using perfusion-corrected 99mTc sestamibi SPECT’, Sci. Rep., 2019/18, 216 [3] McCluskey, S.P., et al. ‘Imaging of Chemotherapy-Induced Acute Cardiotoxicity with 18F-Labeled Lipophilic Cations’, J. Nucl. Med, 2019/60, 1750-1756 [4] Chaher, N., et al. ‘A new collagen III-specific MRI imaging probe to assess cardiac fibrosis’, Heart, 2023/109, 275-276

ABSTRACT. Introduction: Development of radiotherapy-induced lymphopenia (RILP), i.e. a >20% loss in absolute lymphocyte counts, has been observed in approximately half of the treated cancer patients and has been associated with worse response to treatment in the clinic. Although lymphocytes, the key effectors of immunotherapy, are highly radiosensitive, lymphocyte-rich organs are rarely considered as organs at risk during radiation treatment planning. Recent retrospective studies reported associations between the lymphopenia grade and the radiation dose to major blood-carrying structures or thoracic vertebra. However, investigations of the causal relationship between RILP and worse treatment outcome are currently lacking. In this project, we developed murine RILP models and tested whether lymphopenia influenced tumor growth in vivo after radiotherapy and immunotherapy. Materials & Methods: Female C57Bl/6J mice received 10Gy irradiation (2x 5Gy opposed parallel beams at 90 and 270 degrees; 225kV, 13mA) at the heart, large blood vessels (LBV) or thoracic vertebra, using a CT image-guided irradiator, to induce lymphopenia. White blood cell (WBC) subtype counts were measured before and weekly after irradiation. Murine Lewis Lung Carcinoma cells (0.75 million/animal) were inoculated subcutaneously and lymphopenia was induced as indicated above. When tumors reached 208 ± 52mm3, tumors were locally irradiated (10Gy) and animals were injected with L19-IL2 (1 mg/kg, 3 times QOD) or vehicle (PBS) intravenously. Tumor growth was followed via caliper measurements until reaching >4 times starting volume. Results: Irradiation treatment plans were optimized in silico to ensure maximal target coverage, while minimizing the dose to normal tissues. In non-tumor bearing animals, animals receiving LBV (1 target, 3mm circular beam) and thoracic vertebra irradiation (3 targets, 5mm circular beams) developed grade 2 RILP (>20% decrease in lymphocyte counts), which was sustained up to 8 weeks post-irradiation in case of the LBV group. Heart irradiation (1 target, 10mm circular beam) failed to induce grade 2 RILP and therefore was not taken along in the subsequent study with tumor-bearing animals. As previously observed, combination of radiotherapy and L19-IL2 induced a tumor growth delay compared to radiotherapy alone. Grade 2 RILP was confirmed in tumor-bearing animals treated with radiotherapy only, but did not affect treatment outcome. Surprisingly, vertebra and LBV irradiation increase the median survival in animals treated with radiotherapy and immunotherapy (13.8 and 12.6 days for LBV and vertebra irradiation vs 10.5 days in sham lymphopenia irradiation group). Especially in the vertebra group, L19-IL2 combined with radiotherapy restored and increased lymphocyte and eosinophil counts. Further investigations are being performed on the effects of L19-IL2 on the different WBC subtypes. Conclusion: We established in mice an irradiation method to induce minimally grade 2 RILP by precise irradiation of LBV and vertebra. RILP did not affect treatment outcome in animals treated with radiotherapy only. L19-IL2 could be a promising strategy to restore lymphocyte counts and revert RILP.

ABSTRACT. Introduction In radiotherapy of glioblastoma, the clinical target volume (CTV) is relatively large since the individual patient’s tumor extension is unknown. However, accurate geometrical definition of the primary tumor and its microscopic spread has the potential to enhance treatment success. Thus, the objective of this project is to explore biological tumor characteristics in preclinical glioblastoma models for personalized definition of the CTV.

Materials & Methods Two glioblastoma cell lines (U87MG_mCherry; G7-mCherry) were orthotopically transplanted into the brain of NMRI (nu/nu) mice using a stereotactic technique. Tumor growth was monitored with weekly MRI. Treatment (3 Gy/3 fractions, 3 Gy/6 fractions, non-irradiated controls) was performed at 2-3 mm tumor diameter using the in-house developed small animal image-guided radiation therapy system. Brains were excised 72 h after the last fraction and representative tumor slices were analyzed by Matrix-assisted Laser Desorption/Ionization (MALDI) imaging and stained for H&E, Nestin, MMP14, Musashi 1, and CD44, respectively. Another cohort was excised 24 h after irradiation and fresh brain tissue sections were probed using a combined atomic force microscopy (AFM)-light microscopy setup to investigate mechanical properties.

Results The edges of the tumor are clearly shown by the MALDI segmentation via unsupervised clustering of mass spectra and are consistent with the histologically defined border in H&E staining in both models. Specific peaks were identified as potential markers for either normal tissue (e.g. 1339 m/z) or tumor area (e.g. 1562 m/z) by MALDI component analysis. Additionally, 914 m/z showed spatial correlation with MMP14, a stem cell marker, which was positive only for some cells at the tumor margin and thus may represent the invasive front. AFM indentation tests revealed no statistically relevant differences for the mean Young’s moduli of tumors and surrounding tissue in non-irradiated mice but the surrounding was significantly stiffened in irradiated samples compared to the tumor core

Conclusion Our results indicate that an individualized definition of the clinical target volume based on biological tumor characteristics in glioblastoma models seems possible. MALDI imaging is a promising method in this context and correlation with histology-based knowledge is of importance to better understand and validate the hits found in our work. In addition, stiffening of the tissue surrounding the tumor was found after irradiation and will be further investigated by spatial correlation to histological markers. Establishment of a multiplex-staining procedure for a panel of markers related to invasion and extracellular matrix is currently ongoing for these samples.

ABSTRACT. INTRODUCTION: Cachexia is a devastating syndrome characterized by involuntary loss of body weight involving skeletal muscle and adipose tissue wasting. Cachexia occurs in 40-60% of patients with locally advanced (stage III) Non-Small Cell Lung Carcinoma (NSCLC), and is often accelerated during concurrent chemo- and radiotherapy (CCRT), the standard of care therapy for most stage-III patients. Conversely, cachexia contributes to increased treatment-related side effects, which limits the efficacy of current and novel targeted and immune therapies, resulting in poor outcome and survival. Therefore, we aim to delineate the cellular processes and dynamics of tissue wasting during CCRT treatment in an orthotopic lung cancer cachexia (OLCC) mouse model.

METHODS: Immune competent, 11 weeks old male 129S2/Sv mice, were randomly allocated to either sham-operated group or tumor-bearing (TB) groups. Syngeneic lung epithelium-derived adenocarcinoma cells (K-rasG12D; p53R172HΔG) were inoculated into the left lung, and TB mice were allocated to either vehicle treatment (TB) or CCRT treatment (TB-CCRT) starting two weeks post tumor inoculation (average tumor volume: 35 mm3). CCRT treatment consist of three cycles of sub-curative chemotherapy and radiotherapy treatment (once a week). Body weight and food intake were measured daily. At baseline and weekly after surgery, grip strength was measured and tumor growth and muscle volume were assessed using micro cone beam CT imaging. After reaching predefined surrogate survival endpoint, animals were euthanized, and tissues were collected and analyzed for morphological and molecular changes.

RESULTS: Cachectic TB mice showed progressive, concomitant decreases in body weight, fat mass, muscle mass and grip strength compared to sham controls, independent of food intake. TB-CCRT mice showed more pronounced loss of body weight, including muscle and fat mass, compared to TB mice. In contrast to the TB mice, the development of cachexia in the TB-CCRT mice was accompanied by anorexia. In the skeletal muscle of TB mice, markers for proteolysis, both ubiquitin proteasome system (Fbxo32 and Trim63) and autophagy-lysosomal pathway (Gabarapl1 and Bnip3), were significantly upregulated, whereas markers for protein synthesis (relative phosphorylation of Akt, S6 and 4E-BP1) were significantly decreased compared with sham control. The cachectic TB mice exhibited increased pentraxin-2 (P<0.001) and CXCL1/KC (P<0.01) expression levels in blood plasma and increased mRNA expression of IκBα (P<0.05) in skeletal muscle, indicative for the presence of systemic inflammation.

DISCUSSION: We developed an orthotopic mouse model of lung cancer cachexia, which mimics key aspects specific to cachexia in treatment naïve lung cancer patients and during CCRT treatment.

FUNDING: This collaborative project is co-financed with a grant of Danone Nutricia Reseach, a grant of the Nutricia Research Foundation, and PPP allowance made available by Health~Holland, Top Sector Life Sciences & Health to stimulate public-private partnerships.

ABSTRACT. Introduction Radiotherapy for thoracic tumors is often associated with significant radiation doses to heart and lungs frequently leading to damage and side effects. Moreover, growing evidence suggests that patient survival is associated with dose in heart and lungs. As such, interventions are needed to prevent these side-effects. We previously showed in a rat model that heart and lung irradiation can cause fibrotic remodeling of the left ventricle (LV) and vascular remodeling in the lung, resulting in a long-term reduction of cardiopulmonary function. Inhibiting the renin-angiotensin-aldosterone system (RAAS) using an angiotensin-converting-enzyme (ACE) inhibitor reduces LV remodeling, fibrosis and vasoconstriction and is a cornerstone of the treatment of cardiac failure in non-oncological patients. In line with this, in rats RAAS inhibition by Captopril reduced early radiation induced cardiopulmonary damage. However, clinical studies investigating this effect remained inconclusive. In our rat model, combined heart and lung irradiation caused aggravated, persisting effects. Therefore, we hypothesize that heart irradiation potentially prevents protection of the cardiopulmonary system by Captopril. To test this, we investigated the long-term protective effect of captopril after heart and lung irradiation alone or in combination in our rat model.

Methods The heart, 50% of the lateral lung, or both were irradiated. Animals were randomly assigned to receive Captopril (45mg/kg of body weight per day) or normal drinking water (n=11-14 animals per group). LV dimensions, cardiac output and TAPSE were measured using echocardiography. On heart and lung tissue Masson-trichrome and Verhoeff staining were performed to assess collagen deposition and vascular remodeling 38 weeks after irradiation.

Results Cardiac irradiation increased diastolic LV wall thickness from 2.1±0.1 mm to 2.8±0.2 mm (p=0.0004) indicating radiation induced LV remodeling. In line with that, Masson-trichrome staining shows collagen deposition in the LV wall. Heart irradiation reduced cardiac output from 100±3 ml/min to 73±7 ml/min (p=0.0004). Captopril reduced collagen deposition and improved both parameters to 2.0±0.1 mm (p=0.0002) and 92±5 ml/min (p=0.05) respectively. Combined heart and lung irradiation increased LV posterior wall thickness to 2.7±0.2mm (p=0.04) and reduced cardiac output to 70±6 ml/min (p<0.0001). Compared to non-irradiated controls posterior wall thickness and cardiac output did not normalize (2.4±0.2 mm (p<0.0001) and 64±7 ml/min (p<0.0001) respectively) and collagen deposition remained when treated with Captopril. In addition, Verhoeff staining showed vascular remodeling in the lung. This affects the right ventricle, explaining reduced TAPSE (2.1±0.1mm; p<0.0001) compared to control (2.9±0.1mm). Captopril treatment reduced vascular remodeling and improved TAPSE (2.8±0.1mm) if only the heart was irradiated. After combined irradiation, vascular remodeling persisted and TAPSE remained reduced (2.1±0.1mm).

Conclusions In our rat model ACE inhibition using Captopril prevents left ventricular and pulmonary vascular remodeling, and improved associated cardiac function parameters only if the lung is spared. No protection occurred after combined lung and heart irradiation. This may explain negative results of clinical studies that did not consider heart dose when using Captopril, and emphasizes that use of ACE inhibitors for reduction of cardiopulmonary side-effects may require careful patient selection and adapted treatment planning.

ABSTRACT. FLASH radiation is a novel radiotherapy technique that show great potential in improving cancer treatment. However, very little is known about the biological mechanisms behind the highly beneficial FLASH effect. Our research group aims to identify these mechanisms, explain the effect, and to find the optimal way of implementing the technique in clinical practice. For our research, we use in vitro (2D and 3D) models at various controlled oxygen concentrations, as well as in vivo (mice) models. We find that these are highly complementary and both necessary for understanding the FLASH effect. Our studies are performed with a FLASH dedicated 6 MeV electron linear accelerator. Our standard setup involves field sizes from 14-50 mm in diameter (also rectangular field within this range), FLASH dose rates of 2 kGy/s and conventional dose rates of 0.1 Gy/s. A recent update of our setup has also made (6 MV) photon irradiation available, albeit at lower (40-100 Gy/s) FLASH dose rates. Our in vivo mice models are used to investigate the FLASH effect. We have with whole abdominal irradiation of mice shown that the FLASH effect is affected by the pulse structure but mainly depends on the average dose rate or total delivery time. With hemi-thorax irradiation of mice, we have shown that a large FLASH sparing effect is seen (dose modifying factor of 1.5) but also that we get a significant delay in the onset of toxicity following FLASH irradiation. Furthermore, the difference in toxicity between FLASH and conventional irradiation is enhanced by depletion of CD8. We have used various tumour models to investigate treatment response following FLASH irradiation. With subcutaneous models, FLASH tend to be as efficient at reducing tumour growth as conventional irradiation. However, in an orthotopic bladder cancer model we see a reduced treatment effect following FLASH irradiation. More preclinical work is required to fully understand the mechanism behind FLASH and how to fully exploit this new technique clinically.

ABSTRACT. Purpose:

Mini-beam radiotherapy (MBRT) is a novel preclinical approach to radiotherapy (RT), which is based on the use of small parallel beams of radiation (x-ray or protons) to create a highly modulated dose pattern, with regions of high dose (peaks) and low dose (valleys). It has been developed to achieve higher normal tissue sparing compared to conventional RT, while allowing larger doses of radiation.

The aim of this study is to implement a real-time dose minibeam delivery method based on radioluminescence imaging (RLI). First, different scintillating materials are tested in different configurations and then a preclinical setup with a mouse phantom is simulated.

Materials and methods:

MBRT was delivered using a collimator with a thickness of 5 mm and 0.5 mm thick lead slits spaced by 0.5 mm of air, thus giving minibeams with an expected peak-to-peak distance of 1 mm. This collimator was installed on an image guided small animal irradiator (SmART, PXI, 225 kV x-ray source), resulting in a 20 x 20 mm field size.

In order to evaluate RLI for real time MBRT dose measurements different setups were investigated. A scintillator film Gd2O2S:Tb was positioned vertically just below the collimator to evaluate the minibeam shape along the beam axis. The scintillator film was then embedded in four 2.5 mm thick plexiglass slabs to take into account scatter radiation. Two plastic scintillators (Eljen EJ 200) 5x5x5 cm or 5x5x1 cm were also used. The radioluminescence signal was detected using a CMOS camera placed at 90 degrees from the beam axis and the PVDR at the top part of the film/cube was measured. Monte Carlo (MC) simulations were also performed using TOPAS for comparison with the experimental results.

To mimic a preclinical MBRT irradiation setting, a mouse phantom was placed on top of the 5x5x1 cm plastic scintillator slab. Optical images were taken before irradiation (bright field) and during irradiation (RLI). Bright field and RLI images were then superimposed.

Results:

We found that RLI can be used to visualize the preclinical minibeams delivery in real-time. In particular, using the scintillator film alone or the film with plexiglass slabs behind, we measured a PVDR equal to 15 in line with MC simulations. When the plexiglass slabs were placed also in front of the film, the PVDR dropped down to 4 probably because of reflections and scattering of optical photons in the slabs. The same happened with the scintillator cube.

When the mouse phantom was irradiated, superposition of optical and RLI images allows us to clearly see the minibeams on the slab and, thus, to determine the precise location where the mouse was irradiated.

Conclusion:

We proposed an RLI optical method to measure minibeam dose delivery in real-time using radioluminescence imaging. The PVDR values obtained with RLI matches the PVDR obtained with TOPAS MC simulation. RLI of a preclinical irradiation setting allows us to determine the location of minibeams delivery in real-time.

ABSTRACT. Introduction Advancements in radiotherapy techniques are leading to a growing interest in ultra-high dose rate irradiations, both for broad-beam and microbeam irradiations. Synchrotron-based x-ray beams are typically used for microbeam irradiations, since their divergence is very small. Thorough dosimetry testing and the use anatomically detailed small-animal phantoms are therefore necessary to establish irradiation protocols and verify dynamic dose-painting techniques prior to pre-clinical irradiation studies. These phantoms, manufactured with tissue-mimicking materials, can refine or even replace future animal experiments, while providing valuable insights into the accuracy of dose delivery and the implementation of on-line imaging techniques. Materials & Methods This study focuses on the investigation of ultra-high dose rate x-ray fields of the white-beam beamline P61A at the PETRA III synchrotron (DESY, Hamburg, Germany) for future small-animal experiments. This includes the characterization of a 3D printed clear resin (Form 3, Formlabs Inc.), which mainly builds the phantom, together with a bone-equivalent surrogate for the skeleton, and a tissue-equivalent surrogate for mimicking abdomen and brain properties. We measured depth dose curves, to compare this material with those commonly used in dosimetry, by placing GafchromicTM HD-V2-films between plates of clear resin, PMMA, and solid water, respectively. In addition, due to the small beam dimensions (about 2 mm in width), we implemented a dynamic dose-painting technique by translating the set-up vertically (with constant speed) and horizontally through the beam. This allowed for delivering dose to 10-mm wide target volumes in the brain and the abdominal region. For verification, we positioned films in the central sagittal plane of a mouse phantom, as well as at it entrance and exit, respectively. Results The resin used to fabricate the phantom bodies shows x-ray absorption properties that are similar to those of solid water and PMMA, respectively. The experimental depth-dose curves in these materials overlap within less than +/- 10%. Moreover, for an expected dose of (50 +/- 5) Gy at a depth of 5 mm, the average dose to the brain was (38 +/- 4) Gy, while that to the abdomen resulted to be (35 +/- 6) Gy. These two target volumes are at average depths of 6 mm (brain) and 12 mm (abdomen), respectively. When correcting for depth and taking the larger x-ray absorption of the bone surrogate into account, expected and measured dose in the phantom, as well as at its entrance and exit, show a satisfactory agreement. Finally, the dose profiles are uniform within the overall uncertainty, i.e., the horizontal translation of five single beams allowed for an adequate dose coverage of the targets. Conclusion The use of anatomically detailed small-animal phantoms in synchrotron-based radiobiology experiments provides a means to reduce or refine animal studies. A robust quality assurance and a standardized method for end-to-end testing of pre-clinical experiments will accelerate the translation into the clinic, since it enhances the reliability of pre-clinical studies. Future work includes Monte Carlo simulations and the insertion of cell samples into the mouse phantom, made possible by the use of 3D-printed biocompatible boxes.

ABSTRACT. Introduction Recent radiation studies with ultra-high dose rates, UHDR, (>40 Gy/s), known as FLASH radiotherapy (RT), have demonstrated a remarkable reduction in normal tissue toxicity (known as the "FLASH effect") with respect to conventional dose-rate radiotherapy while maintaining similar tumour response. This could represent a step-change in cancer treatment. However, the mechanisms underpinning the FLASH effect are still unknown. Extensive pre-clinical radiation studies are necessary to gain the full understanding of this phenomenon and validate the approach for clinical trials. Xstrahl has developed a pre-clinical FLASH radiotherapy system delivering X-ray beam in UHDR regime (known as SARRP FLASH) to enable research teams to perform precise FLASH investigations. It is essential to develop a methodology to carry out the accurate and traceable dosimetry and have suitable detectors to measure the radiation dose in this regime. The commercially available ionization chambers suffer from significant ion recombination effects when exposed to ultra-high dose rate beams. This makes them unsuitable for dosimetry in FLASH RT. Geometrical constrains also pose challenges for definition of reference conditions and use of conventional dosimetry methodologies which need to be addressed. Xstrahl and NPL secured Analysis for Innovators (A4I) funding to address this measurement challenge. The project commenced in January 2024 and will be delivered in the next six months. Material&methods Alanine dosimetry system will be used as reference standard for the SARRP FLASH. Alanine dosimeters are passive detectors that once exposed to ionizing radiation produce stable free radicals. The number of these radicals is proportional to the total absorbed dose, which can be detected using electron spin resonance spectroscopy. Alanine response is dose rate independent, making it an ideal dosimeter for UHDR beams. Also, alanine pellets can be manufactured in a range of sizes and are routinely used by the NPL established dosimetry service for industrial and clinical applications. However, in the energy range used by the SARRP FLASH irradiator, alanine exhibits significant energy response which require use of a beam quality correction factor (kQ). The kQ factor for alanine will be established using measurements and Monte Carlo (MC) simulations. The project will also develop a suitable measurement jig for SARRP FLASH to enable precise and reproducible dose measurements in these irradiators. Finally, dosimetry guidelines including a full uncertainty budget will be developed to assist users. Results The SARRP FLASH system is already operational in two UK centres. The update on the project progress will be presented during the meeting. Conclusions No current off-the-shelf dosimeters are available to provide accurate dose measurements for pre-clinical photon FLASH RT, and the dosimetry standards for FLASH radiotherapy are still emerging. The cancer research teams need to know accurately the dose delivered to their samples for FLASH research to be valid and reproduced across research groups world-wide. This is also essential to avoid flawed interpretation of the pre-clinical studies, de-risking clinical trials and accelerate clinical adoption of FLASH treatments. The A4I project will deliver a dosimetry solution helping with standardization and accuracy of dose measurements in SARRP FLASH systems.

ABSTRACT. Introduction

Modern preclinical radiobiology research involves a complex chain of events which mimic clinical workflows. As a consequence, there is an increasing awareness of the importance of standardising dosimetry procedures. Dosimetry End-to-End (E2E) tests are designed to determine whether the treatment process leads to the accurate delivery of dose to the intended volumes. These verification procedures play a fundamental role in the success of radiotherapy clinical trials. The development of similar tools for multicentre verification of preclinical radiation treatments would improve benchmarking of laboratories engaged in radiobiological research. Ultimately, this will facilitate a more effective intercomparison of experimental outcomes. We developed End-to-End dosimetry audits based on active and passive detectors with calibration directly traceable to the NPL Air Kerma Primary Standard.

Material&Methods

The first dosimetry audit, E2E_Scint, uses an anatomically correct 3D printed mouse phantom that accommodates the Hyperscint plastic scintillator at two different positions (head and abdomen). The scintillator was extensively characterized (linearity, energy dependence, etc.) and subsequently calibrated against a secondary standard ionization chamber-electrometer system. Measured dose from image guided treatments with different level of complexity, delivered with SARRP, was compared to Muriplan calculated dose. The second developed dosimetry audit (E2E_Ala_Film) combines a 3D printed mouse phantom with alanine at two different positions and EBT3 Gafchromic film, allowing to obtain dose at local points, and 2D dose distributions. NPL alanine dosimetry system is a transfer standard for traceable reference dosimetry. Its energy dependence in medium energy x-rays was previously investigated. Experimental ratios of dose alanine/film were independently validated through Monte Carlo modelling. The system was tested at different beam qualities (Xstrahl 300) for the delivery of non-image-guided treatments.

Results.

The combined standard uncertainty (k=1) of the dose calibration process of the Hyperscint scintillator is 1.5%. The results of applying E2E_Scint over 26 treatment plans (including arcs/obliques-fields/direct/parallel-opposed), showed an average difference between the measured and the Muriplan calculated dose of 1.1%. The maximum difference was 3.3% for a brain plan combining oblique-fields with the 3 mm x 3 mm collimator. While testing E2E_Ala_Film for the verification of total body irradiation treatments to the mouse phantom, the average of the difference between the dose measured with film and alanine was -0.3%. The combined standard uncertainty (k=1) for the determination of the differences among irradiations with beam qualities between 0.5 and 3.5 mm was 3.2%.

Conclusions.

The combination of passive and active detectors with anatomically correct mouse phantoms are adequate for the development of End-to-End dosimetry audits for the independent verification of preclinical radiation treatments. The traceability of the detectors’ calibration to primary standards strengthens the dosimetry chain in the validation of preclinical plans, and it is consistent with the current practice for dose traceability of clinical radiotherapy treatments. Dosimetry audits are an important tool to improve quality of reported results and to support standardization of preclinical radiation research. Their implementation at national and regional levels could lead to databases of anonymised records, which will positively impact the dissemination of best practices and sharing of validated results.

ABSTRACT. Introduction: To ensure a high reproducibility and accuracy in pre-clinical experiments, it is necessary to implement adequate training programs for the use of small-animal irradiation infrastructure. Additionally, ethical considerations, particularly the ‘3R’ principles in animal research (Replacement, Reduction and Refinement), must be taken into account. Anatomically detailed mouse phantoms play a major role in all these aspects. In addition, use of precise multi-modal imaging is increasing and phantoms that reflect anatomic details are essential in order to establish image registration and positioning procedures, as well as for dosimetry purposes. Materials & Methods: We modified a previously developed 3D-printed modular mouse phantom (Wegner et al. doi: 10.1088/1361-6560/acc566.) to accommodate a PTW microdiamond detector, GafchromicTM EBT3 films, and thermoluminescence dosimeters in the head and abdominal region. The procedures developed at the production site of the mouse phantom, which is equipped with a small-animal irradiation platform (SmART+, with integrated CT imaging and the Pilot® treatment planning system) were tested in a multi-center end-to-end test setting. The pre-clinical irradiation protocol followed a different workflow, comprising off-line CT imaging (X-CUBE, Molecubes), the µ-RayStation (RaySearch Laboratories) treatment planning system (TPS), and a custom x-ray irradiator (YXLON Maxishot, YXLON International GmbH). Treatment plans with a mean prescribed dose of 2 Gy were created for the head and abdomen with one lateral beam collimated to 5, 10 and 20 mm. Dose values obtained with the microdiamond detector (µD) were compared to the planned dose, and complemented by a gamma analysis (3%/1 mm) based on EBT3 film dosimetry and TPS µ-RayStation. Results: Measurements with the µD detector in both the head and abdomen revealed acceptable accuracy for 10 and 20 mm field sizes with maximum deviations of 3% from the TPS. For a field size diameter of 5mm deviations of up to 12% were observed. Gamma analysis resulted in passing rates of 80% and 48% for a 20 mm and 5 mm field size in the abdomen, respectively. The latter for the smallest field size reflect the high sensitivity of the positioning procedures. Conclusion: We demonstrated the versatility of a robust 3D-printed mouse phantom, which can be sent via ordinary mail and adapted to accommodate various dosimeters, depending on the users’ needs. An end-to-end test of cerebral and abdominal irradiations with x-rays revealed the need to improve the phantom positioning accuracy when irradiating with fields below 10 mm diameter. In a next step, the usability of the mouse phantom for an end-to-end test in particle beam research will be tested

ABSTRACT. Introduction: Computed tomography (CT) imaging is a key component of image-guided radiotherapy (IGRT) in the clinic and the laboratory using small animal irradiators. Similar to advances in the clinic, quantitative image analysis using radiomics can be used to gain information about the underlying biological characteristics of tumours and normal tissues towards the development of imaging biomarkers for precision medicine applications in radiation oncology. In this study, we conducted a cross-centre comparison of reliable radiomics features derived from scans obtained using a dedicated micro-CT (µCT) scanner and cone-beam CT (CBCT) scanner on the small animal radiation research platform (SARRP) aiming to develop comprehensive and transferable radiomics signatures. Materials/Methods: µCT and CBCT scans of a phantom and two mouse models were acquired using a Quantum GX2 (Revvity, UK) and SARRP (Xstrahl), respectively. Different volumes (phantom) and tissue densities (mouse) were segmented from the scans and the reliability of radiomics outputs was assessed using different imaging protocols and harmonisation of pre-processing parameters. Reliability was measured through intraclass correlation coefficient (ICC) analysis. Results: The reliability of in phantom radiomics features differed across segmentation volumes, with first order and GLCM features being the most stable feature types identified across scanners and volumes. Overall, µCT imaging produced more reliable features compared to CBCT in mice with notable superiority in higher-density tissue (bone). Tissue density-specific preclinical radiomics signatures were developed for the lung (133 features), heart (35 features), and bone (15 features) that were shared across CBCT and µCT modalities. Variations in the reliability of radiomics features across scanners were observed yet normalisation steps including standardisation of imaging energy and pre-processing factors (voxel size) can be used to allow accurate comparisons across µCT and CBCT scans. Conclusions: µCT and CBCT scans can be used for radiomics analysis to gain a better understanding of the underlying biology of tumours and normal tissues. This study demonstrates the importance of standardisation and emphasises the need for multi-centre radiomics studies to assess feature reliability to aid widespread application and ultimately the clinical integration of radiomics.

ABSTRACT. Radiotherapy remains one of the most effective modalities for anticancer treatment and part of standard of care for many primary and metastatic brain tumors. Boosting the efficacy of radiotherapy is therefore a logical avenue to improve patient survival. We have developed a radiosensitization strategy called ‘induction of mitotic enrichment’. It has long been known that the radiosensitivity of a cell depends on the phase of the cell cycle and that mitotic cells are especially vulnerable. Enriching the tumor for mitotic cells by arresting them during division prior to each radiotherapy fraction should therefore render the tumor population more sensitive to irradiation. Ideally, induction of mitotic enrichment should be reversible and non-cytotoxic to prevent healthy tissue toxicity and be compatible with clinically applied fractionated radiotherapy regimens. We have now identified an orally available targeted tubulin polymerization inhibitor that can achieve repeated and reversible mitotic enrichment for up to 10 hours prior to radiotherapy, without causing cytotoxicity in vitro or healthy tissue toxicity in vivo. This tubulin inhibitor (ABT-751) efficiently radiosensitizes a range of preclinical tumor models representing adult and pediatric, primary and metastatic brain tumors. Importantly, ABT-751 also improves survival in mouse models of adult glioblastoma (GBM), but only in a mitotic enrichment setup when given several hours prior to radiotherapy to allow accumulation in mitosis. Of note, induction of mitotic enrichment does not exacerbate healthy tissue toxicity. To conclude, induction of mitotic enrichment by ABT-751 radiosensitizes adult GBM to fractionated radiotherapy. We are currently expanding our preclinical development of mitotic enrichment as a radiosensitization strategy to other targets and cancers and have designed a phase 0/I trial for adult GBM to demonstrate induction of mitotic enrichment and tolerability when combined with radiotherapy. We expect to start enrollment in this trial soon.

ABSTRACT. Introduction: In our earlier work, Rezaeifar et al. (DOI 10.1088/1361-6560/ace308), a 3D U-Net deep learning model was used to capture the information from preclinical bioluminescence imaging (BLI) and predict the location of luciferase-enabled orthotopic glioblastoma tumor model in rats. In this method, a Monte Carlo simulation (MCS) engine was used to track the emitted photons up to the skin of the animal, simulating the bioluminescence skin fluence (BSF), using the geometry derived from a set of contrast-enhanced cone-beam CT (CE-CBCT) images. Thereafter, a training database including 42 cases was constructed comprising MCS of BSF for real animal cases. The deep-learning solution then took the simulated BSF data and predicted the tumor location and shape with a median dice score of 61% and therefore enabled BLI-based tumor targeting for each radiation fraction for glioblastoma. This study aims to improve the performance of the previously provided solution, hereafter called BLI-only approach, by incorporating anatomical data obtained from the cone-beam CT (CBCT) scan, hereafter called CT-BSF approach.

Materials and Methods: In this study, a 3D multi-channel UNet model is employed to simultaneously take the CT and BSF data as input. A specific normalization approach is defined to ensure that both channels of information are in the same order of magnitude when presented to the AI model, to avoid input bias for specific channel. A hyper-parameter optimization algorithm was applied to obtain the best set of hyper-parameters for the model ensuring an optimized solution. The predicted tumor location was then scored against the ground truth CE-CBCT-based tumor contour, delineated by an expert biologist, and compared against the BLI-only approach and a CT-only counterpart model. Furthermore, the performance and robustness of the CT-BSF approach was compared to the CT-only model to investigate the added benefit of the BLI data. For this purpose, the effect of artificially added noise in each input channel is investigated.

Results: The CT-BSF approach exceeds the performance of the CT and BLI-only models measured by average dice score over the same test database, with dice scores of 83 ± 20%, 81 ± 18%, and 61 ± 17% respectively. The robustness analysis shows that the CT-BSF model is greatly resistant against the input noise modelled by reducing number of particles in the MCS. Furthermore, for the added Gaussian noise in the CT images, the CT-BSF model shows superior robustness against the CT-only model.

Discussion: The proposed solution in this study shows a performance gain from utilizing complementary information by a multichannel AI solution to improve the prediction accuracy of BLI-based tumor targeting. However, the results show that the majority of the gain is due to the tumor appearance in the Contrast Enhanced CT images which can be seen from a modest gain of only 2% with the addition of the BLI input channel. Nevertheless, the results showed an improved robustness while dealing with the image noise in the CT images which underlines the usage of both input channels in the CT-BSF model.

Funding: This work was partially funded by a double PhD program with the University of Hasselt and Maastricht University.

ABSTRACT. Introduction: Medulloblastoma (MB) is the most frequent high-grade brain tumour of childhood and adolescence and an important contributor to brain tumour-related mortality and morbidity. Multimodal therapy of MB consists of combined surgery, chemotherapy, and radiotherapy (RT), with RT being delivered to the entire central nervous system (CNS). This leads, besides to other late effects, to pronounced neurocognitive impairment. New treatment strategies to improve tumour control and reduce toxicity are being currently developed in our institute at the pre-clinical level and include the use of radiosensitizers, thus requiring the implementation of a precise cranio-spinal irradiation (CSI). For optimising the workflow, we performed plan studies and dosimetric verifications with the help of mouse phantoms. Materials & Methods: Future pre-clinical CSI will be carried out with a SmART+ platform, equipped with various fixed collimators to form rectangular, quadratic, or circular fields. Mice will be placed inside a cylindrical anaesthesia unit, which allows for monitoring vital parameters. With the help of additive manufacturing, we produced two anatomically detailed mouse phantoms, with two typical spine shapes. To mimic typical irradiation settings, we carried out a CT scan of the mouse phantoms inside the anaesthesia unit, contoured planning target volume (PTV) and organs at risk (OARs), proceeded with treatment planning, and concluded with the irradiation. The PTV, which should receive a fraction dose of 2 Gy, was obtained by expanding the contours of the CNS by 2 mm. Femoral bones, lungs, and abdomen were defined as OARs. We compared different treatment plans, obtained by using three to six isocenters and set-ups comprising opposing lateral fields, alone or in combination with a dorsal field. We measured the delivered dose in the sagittal plane at the isocenter position with GafchromicTM EBT3 films. Results: A mean dose of (2.1 +/- 0.1) Gy to the PTV could be obtained with all treatment plan variants, independent of the spine shape. A straighter spine allows for the use of fewer isocenters and field sizes. Irradiation geometries with lateral opposing field are suitable to spare thorax and abdomen, e.g., with mean doses between 0.6 and 0.7 Gy and D20% between 0.5 and 1.5 Gy, respectively. The use of a dorsal field for the caudal region allowed us to further spare the femural bones. The mean dose value, for example, could be reduced from about 2.5 Gy to about 0.5 Gy, while D20% dropped from 5 Gy to 0.5 Gy. The measured dose distribution was in good agreement with the calculated one, with a maximum relative dose increase of 25% in the regions where the radiation fields overlap. Conclusion: The use of mouse phantoms allowed us to optimise precise CSI prior to implementation in mice models of MB. With enough training, the total treatment time could be kept between 30 and 40 minutes, depending on the spine shape and the number of used isocenters, with a satisfactory PTV coverage and sparing of normal tissue. Future work will include the analysis of possible mouse movement to improve field overlap without the risk of PTV underdosage.

ABSTRACT. Introduction: Radiotherapy (RT) is a commonly used treatment option for a range of cancers, typically in combination with surgery and/or chemotherapy. While effective in controlling or eliminating localized tumors, RT induces a host of normal tissue toxicities. Gastrointestinal (GI) side effects from RT treatment are very common; often, limiting the RT dose that can be delivered to GI tumors and, thereby, its effectiveness. Similarly, RT-mediated cardiovascular disease is thought to occur through both direct and indirect tissue damage, i.e., through incidental dose delivery to the heart and through the effects of dose deposition to surrounding lung tissue. Here we present the development of two clinically relevant, image-guided mouse models of RT-induced tissue toxicities, using the Small Animal Radiation Research Platform (SARRP, Xstrahl). Materials and Methods: Histologically, γ-H2AX, EdU incorporation, TUNEL assay, H&E, and Trichrome staining were used to characterize both acute and chronic effects of local irradiation. Radiologically, quantitative echocardiography, 18F-fluorodeoxyglucose positron emission tomography-computed tomography, and perfusion imaging with Technicium-99 sestamibi were implemented to identify sensitive imagining measures of cardiac injury. Hematologically, multiple cytokines were measured to establish the levels of injury and inflammation. Results: Our findings suggest that our image-guided RT injury mouse models are feasible since we observed increased levels of DNA damage, apoptosis, inflammation, and fibrosis in the focally irradiated areas compared to the non-irradiated adjacent areas. Moreover, we found significantly increased levels of multiple pro-inflammatory cytokines and chemokines in the irradiated tissues and plasma samples. Conclusions: We developed and characterized two pathophysiologically relevant mouse models of radiation-induced GI and cardiac toxicities. Both models constitute new tools to rapidly evaluate multiple biological markers of acute and chronic toxicities and may be used to integrate radiomic and biochemical markers of toxicity to inform early therapeutic interventions for human translational studies.

ABSTRACT. Parallel to the start of proton radiotherapy in 2013 preclinical experiments were established in the experimental area of the University Proton Therapy Dresden. Embedded in the OncoRay research building, experimenter’s of the proton area have access to cell laboratories and an animal facility including a preclinical imaging platform. First, a double-scattering setup was developed creating a 150 MeV spread-out Bragg peak (SOBP) for in-vitro studies within a homogeneous 10 x 10 cm² field (1). The inclusion of suitable collimators and range shifters in the beam path together with simple optical imaging and positioning enabled the irradiation of subcutaneous tumours on mice hind legs (2) and ears (3). In this process, a mouse bedding unit was developed allowing for CT and MRI imaging, photon and proton treatment without positional changes of the animal and preventing contamination at the same time (2). In a next step, a setup for partial brain irradiation of mice was developed making use of proton radiography for positioning at the beam – firstly by landmark-based co-registration of the bedding unit (4) and later on by target delineation directly on a proton radiography of the skull (5). A 90 MeV pristine proton beam was laterally collimated by 3 mm and 4 mm apertures and decelerated such that it stops in the middle of the mouse head. Irradiated mouse brains of two strains of differing radiosensitivity allowed insights into acute radiation damage, e.g. by double-strand break analysis (4), relative to photon irradiation. Moreover, late endpoints like radiation induced contrast enhancement were studied by regular MRI follow up over six months and subsequent histological analysis (6). The correlation of MRI imaging, spatial histology information and Monte Carlo dose- and LET-simulations provide further insights into the radiobiological response to proton irradiation. To develop this setup further a first prototype small animal irradiation device (SMART IB, Precision X-ray) was installed at the proton beam line. The device is currently commissioned and will allow on-site CT-based treatment planning and positioning as well as photon irradiation for reference. Lately, the Flash irradiation of mouse brains with protons was realized combining a dedicated 3D-range modulator (7) to produce an SOBP with one pencil beam and proton radiography for accurate mouse positioning. Conventional and Flash irradiation of six C57BL/6 each were performed and the animals were followed for 6 months; whereby analysis is still ongoing. To sum up, during the last years preclinical experiments with protons were developed starting from simple adaptations of an existing double-scattering setup to dedicated setups for partial brain irradiation with proton Flash. The integration of the small animal irradiation device will further enhance the experimental possibilities bringing the preclinical animal studies one-step closer to clinical-like treatment.

References: (1) Helmbrecht et al. (2016) J Instrum (2) Müller et al. (2020) Biomed Phys Eng Express (3) Kroll et al. (2022) Nat Phys (4) Suckert et al. (2020) Radiother Oncol (5) Schneider et al. (2022) Front Oncol (6) Suckert et al. (2021) Front Oncol (7) Horst et al. (2023) Front Phys

ABSTRACT. Introduction

Over the past decade, several small animal irradiation platforms with integrated image guidance have been developed and commercialized for X-rays. Conversely, for protons and light ions, only a few small animal irradiators have meanwhile been developed at selected centers, typically limited to the combination of passively scattered beams with the same X-ray cone beam computed tomography imagers developed for the abovementioned X-ray radiation therapy platforms. This setting, however, limits the achievable beam quality and targeting accuracy and does not exploit the unique image guidance possibilities of ion beams, besides implementing a delivery scheme progressively phased out in clinical practice. Hence, this work aimed to fill this research gap by providing a novel platform for small animal proton radiation research in line with clinical standards.

Materials&Methods

In the ERC (European Research Council)-funded project SIRMIO (Small animal proton irradiator for research in molecular image-guided radiation-oncology, www.lmu.de/sirmio) we developed an innovative portable system to enable precision image-guided small animal proton irradiation at clinical proton therapy facilities. The modular system combines a dedicated beamline consisting of passive (degraders/collimators) and active (quadrupole magnets) components for precise dose application with advanced image guidance specific to proton therapy, with different implementations of proton radiography/tomography along with in-situ in-vivo verification of the actual treatment delivery with positron emission tomography and (for pulsed beams) ionoacoustics. For treatment planning we rely on a validated research system from RaySearch Laboratories, tailored to reproduce the SIRMIO beams.

Results

In the first experimental characterization and commissioning at the Danish Center for Proton Therapy (DCPT) we demonstrated the ability of the SIRMIO platform to degrade and actively focus an incoming clinical proton beam and perform automated, precision, image-guided delivery to homogenous and heterogeneous, mice-mimicking phantoms. Examples highlighting the imaging and treatment workflows implemented in SIRMIO, with emphasis on the main capabilities of flexible and accurate dose delivery with on-board image guidance, will be presented.

Conclusion

The SIRMIO platform can provide a versatile and powerful tool for precision, image-guided radiation research with proton beams. First application to experiments with mice is being planned and is foreseen to be performed at DCPT within the course of 2024.

Acknowledgement: This project has been supported by the European Research Council (grant agreement 725539), and EU projects 730983 (INSPIRE) and 101008548 (HITRIplus) for transnational access. We also acknowledge support from DFG (grant agreements 299102935, 372393016, 455550444) and from the Center for Advanced Laser Applications (CALA), and thank the broad network of collaborators, particularly the DCPT team, E. Traneus and R. Nilsson from RaySearch Laboratories AB, C. Granja and C. Oancea from Advacam, F. Becker from Vacuumschmelze GmbH, H. Kang and T. Yamaya from NIRS-QST, A. Zoglauer from University of California, Berkeley, and J. Gordon from Pyramid Technical Consultants. N. Bassler is supported by NovoNordisk Foundation (grant number NNF195A0059372) and the Danish Cancer Society (grant no. R191-A11526).

ABSTRACT. Introduction

For image-guided small animal irradiations, the preclinical workflow (imaging, contouring, irradiation planning, and delivery) is preferably performed in a single session. As prolonged exposure to anaesthesia may impact the outcome of the experiment and and animal well-being, it is important to speed up the workflow as much as possible. One aspect where time can be gained is the dose computation. Monte Carlo (MC) simulation is the gold standard for preclinical dose calculations. MC-based calculations are, however, rather time-consuming because a large number of particles is needed to achieve sufficient statistical accuracy especially since animal CTs are comprised of very small voxels (≤100 μm). An alternative is to use a deterministic dose algorithm, which is more computationally efficient [1]. In this work, we investigate the dose calculation accuracy of a deterministic dose engine for small animal proton irradiations.

Materials & Methods

We approximate the solution to the Linear Boltzmann transport equation (LBTE), which describes the proton phase space density. The LBTE is simplified into two partial differential equations: the one-dimensional Fokker Planck (FP) and the Fermi-Eyges (FE) equations, which are solved numerically using the discontinuous Galerkin method and analytically (for a Gaussian boundary condition) on a discretized three-dimensional grid, respectively.

To properly model the dose in the presence of heterogeneities, the beam was decomposed into 25 smaller beamlets arranged in concentric circles with optimized weights, distances, and sizes. The splitting scheme used is 1+6+6+12, which represents 1 central beam and 24 beamlets distributed on three concentric circles wherein 6 beamlets are placed on the innermost ring and 12 beamlets are placed on the outermost ring. A treatment plan for a 2 mm spherical target in the brain of the MOBY phantom [2] was generated in matRad [3]. The plan was optimized to deliver 9.5 Gy to 98% of the planning target volume. The dose distribution based on this plan was calculated using the deterministic code and was compared to results from TOPAS MC simulations for the same setup.

Results

The dose distributions from the deterministic code and TOPAS were compared. Good agreement was observed in terms of the proton range wherein the difference is less than 0.1 mm (i.e., size of the dose calculation grid). Gamma analysis using 3%/0.1 mm criteria revealed a 99.3% passing rate, which shows that there was no major dosimetric difference with TOPAS. Furthermore, the deterministic code was able to calculate the full plan with 111 spots in 3 min, while TOPAS took 25 min to do the same job in the same system.

Conclusion

Our findings show that the deterministic algorithm can achieve the same level of accuracy as the gold standard MC codes with significantly less time. The outcome suggests that the deterministic code is a suitable alternative for a more efficient treatment planning in small animals.

References

[1] Tiberiu Burlacu et al. (2023) https://doi.org/10.1080/23324309.2023.2166077 [2] William Segars et al. (2004) https://doi.org/10.1016/j.mibio.2004.03.002 [3] Hans-Peter Wieser et al. (2017) https://doi.org/10.1002/mp.12251

ABSTRACT. Introduction: A control system was developed for interfacing with the XStrahl Small Animal Radiation Research Platform (SARRP) and proton beam control software, to allow for novel clinical experimentation methods to simulate proton Pencil Beam Scanning (PBS) with a fixed experimental beam line at conventional and FLASH spot dose rates.

Materials and Methods: The control software consists of a Python-based program (pySARRP) and a LabVIEW interface (viPBeam), which employ the TCP/IP communication protocol for sending and receiving position and beam control commands. The viPBeam control software is designed to take a user's input containing relevant parameters such as beam spot positions, map size and spot spacing, desired monitor units, dose rate, and threshold, which are then used to communicate with pySARRP and the proton beam control software, to control and automate the initialization of SARRP, the positioning of the SARRP robotic stage, and the proton beam state and delivery. The positions of the robotic stage create the desired scan pattern necessary to simulate proton PBS. pySARRP and viPBeam are equipped with safety and limit checks, and a graphical user interface for control and feedback. We use a fixed angle beam line in an allocated research room to deliver a 230 MeV proton beam with a range of 32 g/cm2 from an IBA Proteus Plus proton therapy system. The nominal beam current ranges between 2 nA to 360 nA , for conventional and FLASH, respectively. We validate the newly developed programs (pySARRP and viPBEAM) by controlling the robotic stage of the SARRP to emulate PBS proton radiotherapy for the fixed beam line. Validation experiments were done using raster maps 2x2, 3x3, and 4x4 with 1mm spot spacing to measure the latency time.

Results: Several tests were successfully carried out using the newly developed control software with a 4x4 raster pattern and the proton beam operated at both conventional and FLASH dose rates. To ensure correct beam delivery, a radiochromic film was used to measure and display the programmed/delivered raster pattern. We irradiate each map six times with stage acceleration of 30 mm/s2 and velocity of 20 mm/s. The measured transition time between spots is 1.47 ± 0.175 s, which includes 1s of the stage motion time. While the transition time is not considered in the definition of instantaneous dose rate, the average dose rae and the PBS dose rate include it in their definitions.

Conclusions: An easy-to-use control software was developed and implemented to handle communication and real-time control of a SARRP and a proton beam for carrying out automated proton PBS FLASH experiments. The developed control program reliably delivers the PBS beam in FLASH and conventional modalities for small animal radiobiology experiments.

ABSTRACT. Introduction Small target volumes at shallow depths in combination with the path length sensitivity of charged particles pose an inherent challenge for pre-clinical proton irradiation. Precise and reproducible positioning throughout the whole workflow is crucial to achieve the accuracy necessary to correlate dosimetric parameters with morphological changes. This work aimed to assess the influence of positioning inaccuracies on the dose and dose-averaged linear energy transfer (LETd) distributions.

Materials and Methods 13 BALB/cJRj mice received two spiral cone-beam µCT images (Molecubes, Belgium, 0.225 mA, 50 kV, slice thickness 0.2 mm) in prone position for treatment planning (pCT) and verification (cCT). cCTs were acquired to estimate the repositioning accuracy during irradiation. Throughout the workflow the anaesthetised mice were immobilized with an in-house designed bedding unit including a toothbar and landmarks. Rigid image registration was performed between the cCT to pCT based on these landmarks, which were utilized for positioning during irradiation employing a robotic couch and an in-room laser system. All mice were irradiated with a lateral horizontal proton beam with a diameter of 5 mm (Knäusl et. al. 10.1016/J.EJMP.2023.102659). A median dose of 66 Gy(RBE) was prescribed to a cylinder (Ø3 mm, 4 mm length) encompassing the right hippocampal region. Dose and LETd calculation was performed in RayStation (v14.01.110.0, MC dose engine v5.5, resolution 0.2×0.2×0.2 mm3). Treatment plans (6 spots with energies ranging from 69.5-74.2 MeV) were optimised concerning target coverage and sparing of the contralateral brain hemisphere. All treatment plans were recalculated on the cCTs and median changes (including range) in dose and LETd values were evaluated in the target region and contralateral brain for all 13 mice. Moreover, D10% and D90% were reported as surrogates for the maximum and minimum dose in these very small volumes together with the homogeneity index HI ([D10%-D90%]/D50%).

Results Recalculation of the delivered dose on the cCT showed that deviations in median D50% and D10% values were within 1 %. The median D90% dropped from 64.3 (min 63.9, max 64.8) Gy(RBE) to 57.1 (min 13.0, max 64.0) Gy(RBE). The median HI increased from 0.04 to 0.15. V95% decreased from 95.7 (min 95.0, max 98.0) % to 75.0 (min 23.0, max 94.5) %. The LETd in the target was not significantly affected by the recalculation on the cCT (median LETd90% and LETd10% deviations within 1 %). For the contralateral brain D10% decreased from 11.5 (min 4.9, max 21.0) Gy(RBE) to 3.9 (min 0.9, max 24.0) Gy(RBE).

Conclusion The reproducibility evaluation showed comparable D10% values in the target and surrounding tissue, indicating that the estimated positioning uncertainties do not pose additional harm to the animals. For most animals the dose reduction to the contralateral brain resulted from decreased target coverage. A potential underdosage of the target was identified, in particular for two outliers, and reflected in the D90%. The small deviations in LETd parameters were expected, as the sagittal LETd distribution was relatively homogenous. An improved positioning procedure is required to ensure sufficient target coverage and preliminary analysis of a couch-mounted (clinical) CBCT showed promising results.

ABSTRACT. Introduction Clinical research indicates that late side-effects after cranial proton irradiation may manifest in some patients due to a variable relative biological effectiveness (RBE) and sensitivity of the ventricular proximity. To investigate these parameters, we opted for a preclinical model, utilizing neuroinflammation as the endpoint, given its pivotal role in the brain's radiation response. Microglia, the brain's resident immune cells, exhibit increased proliferation and activation after injury. Therefore, we quantified neuroinflammation based on the morphological changes that activated microglia undergo.

Materials & Methods Microglia proliferation and activation at cell specific level are analyzed following proton irradiation of the right hippocampi in C57BL/6 mice. The cohort includes control (n = 1) and mice irradiated with 45 Gy (n = 3), 65 Gy (n = 3), and 85 Gy (n = 3). After treatment, mice underwent 6 months of longitudinal magnetic resonance imaging (MRI). Excised brains were axially sliced into approximately 30 sections per brain (3 µm thickness, 100 µm intervals) and stained for Ki67, Iba1, and DAPI to detect proliferation, microglia, and nuclei. An in-house developed algorithm segmented microglia, providing information on their ventricular distance and morphological parameters for calculating the M-Score, i.e. activation status. Ki67-positive microglia were identified with an established Ilastik classifier, and Slice2Volume was used to co-register immunofluorescence images to planning CT data on which proton dose and linear energy transfer (LET) distributions were calculated using Monte Carlo simulations.

Results Population-based analyses in whole mouse brains revealed a dose-dependent increase in microglia proliferation and activation. Current spatial findings indicate a gradual increase in microglia activation status (i.e. M-score) within the isodoses of 45 Gy irradiated mice and a more pronounced escalation for those exposed to higher doses. Microglia with a high M-score are more prevalent in the periventricular region, the dose maximum area, and the distal beam edge. The correlation between microglia activation and ventricular distance appears more prominent in mice irradiated with 65 Gy than 45 Gy. Initial results indicate that Ki67-positive activated microglia in 65 Gy irradiated mice are broadly distributed throughout the irradiated area, not limited to the highest microglia density region. In contrast, 45 Gy irradiated mice show few Ki67-positive activated microglia, evenly distributed across irradiated and non-irradiated brain regions. Consistent with the proliferation increase with higher doses, microglia density in middle brain slices of mice irradiated with 65 and 85 Gy is approximately 120 cells/mm² and 184 cells/mm², respectively, compared to the control and 45 Gy irradiated mice, which have similar densities at around 80 cells/mm². Interestingly, the rise in microglia density corresponds to the increasing contrast enhancement leakage area in the final MRI for all mice.

Conclusion We performed section-based histo-cytometry to investigate cell specific response in relation to spatial parameters. Findings show that proliferation and activation of microglia is dose- and location-dependent. So far, our data supports microglia inflammation as potential biomarker for radiation-induced brain damage. Further correlations of microglia activation with dose versus dosexLET are ongoing.

ABSTRACT. Introduction

Similar to the clinic, the preclinical workflow involves imaging, contouring, irradiation planning, and delivery. For image-guided small animal irradiation, this entire process is typically performed in a single session during which the animal is maintained under anaesthesia. To avoid prolonged exposure to anaesthesia and reduce the workload, efforts are made to achieve a faster workflow by reducing the contouring time. Deep learning auto-contouring models for small animals have shown promising results with millisecond inference times. However, their impact on the dose distribution has not been demonstrated yet. In this work, we assessed the dosimetric impact of deep learning-based segmentation models on animals.

Materials & Methods

Two deep learning models, nnU-Net [1] and μ-RayStation (RaySearch Laboratories AB, Stockholm, Sweden) [2-3], were trained on 105 micro-CT scans of mice with manual delineations of the spleen, lungs, heart, liver, intestine, bladder, and kidney [4]. The performance of the trained networks was assessed on an independent dataset (N=35) excluded from the training. For geometric evaluation, the automated contours were compared with the manual contours (ground truth) using the Dice similarity coefficient (DSC) and 95th percentile Hausdorff distance (95p HD). For dosimetric evaluation, proton treatment plans for the spleen were generated in μ-RayStation and dose distributions were calculated on a 0.3 mm isotropic dose grid. The dosimetric impact was evaluated by performing 3D gamma analysis (3%/0.3mm) between the plans based on the automated (nnU-Net and μ-RayStation) contours and the original plan based on the manual contour. Furthermore, the plans based on the automated contours were assessed on whether they still meet the dose coverage criteria of D98 ≥ 95% for the manual contour.

Results

Both deep learning models provide similar results in terms of the average DSC scores (nnU-Net: 0.89 ± 0.07, μ-RayStation: 0.87 ± 0.07) with the spleen showing the lowest overlap at DSC ≈ 0.7. The average 95p HD were 0.63 ± 0.34 and 0.69 ± 0.35 for nnU-Net and μ-RayStation, respectively. The largest segmentation errors (> 1 mm) were observed for the liver and intestine. Gamma analysis using 3%/0.3 mm criteria revealed an average of 52% ± 14% and 48% ± 12% passing rate for nnU-Net and μ-RayStation, respectively, which suggests that there were dosimetric differences compared to the plans based on the manual contours. The average D98 were 8.9 ± 1.4 Gy and 8.1 ± 1.1 Gy for nnU-Net and μ-RayStation, respectively.

Conclusion

We demonstrated the feasibility of two different deep learning models for segmentation in mice. Both models perform similarly and correlate well with each other. Dose distributions were calculated on the spleen, which is one of the most difficult organs to delineate. The poor agreement between the automated contours and the manual contour for this organ is reflected on the quality of the dose distributions that can be achieved. Work is in progress to elucidate the large differences between automated and manual contours.

References [1] Isensee, F. et al. https://doi.org/10.1038/s41592-020-01008-z [2] Ronneberg, O. et al. https://doi.org/10.1007/978-3-319-24574-4_29 [3] Çiçek, Ö. et al. https://doi.org/10.1007/978-3-319-46723-8_49 [4] Rosenhain, S. et al. https://doi.org/10.1038/sdata.2018.294

ABSTRACT. Introduction The next generation of preclinical radiation biology experiments particle beams are demanding on the control of the delivered dose distributions. The anatomical variations of tumor and normal tissues require image guided planning to obtain high conformity between desired and delivered dose distributions. Therefore, a new beam line for image-guided particles irradiations equipped with an on-line X-ray CBCT has been developed.

Materials and methods The PARTREC cyclotron offers particle beams (H, He, C) with variable energy for small animal irradiations. Both shoot-through irradiations and spread-out Bragg peak irradiations are performed. Depending on experiment requirements scattered or scanned beams are used. FLASH dose rates are under development for proton and helium beams. The Geant4-based simulation packages BDSIM and TOPAS and measured beam parameters were used to optimize the beamline and the final beam shaping. The constraints imposed by the on-line CBCT were integrated in the optimization.

Results The beam line layout for shoot through experiments is the same as that of the existing beam line. The newly designed scatter system produces a flat field (≤2% homogeneity) with an efficiency ≥30 % for a field size ≤70 mm, thus enabling FLASH dose rates. Uniform scanning is also done with this beam line layout. A dedicated collimator close to the irradiation position inside the CBCT shapes the field laterally. For spread-out Bragg peak pencil beam irradiations, a small spot size at the position of the final collimator is obtained with a telescopic beam optics. The beam path in air has been minimized to obtain a good transmission. To achieve this the vacuum of the beam line is extended into the CBCT to 200 mm from the irradiation position during irradiation. Also, the distance between the final collimator and the irradiation position has been minimized to obtain the smallest possible lateral penumbra. With this configuration a transmission of 25 % and a 0.35 mm (20 - 80 %) penumbra at the irradiation position in air can be achieved according to the simulations, thus enabling FLASH dose rates. Scanning is performed by moving the animal. Putting the range shifter and modulator up- or downstream of the collimator has an opposite effect on transmission and lateral penumbra and will thus be determined by the requirements of the experiment. The simulations show that the alignment between beam and collimator axes is critical for beam shape and transmission. Therefore, two 2D steering magnets have been installed to optimize beam position and angle at the collimator entrance. At the primary energies used for spread out Bragg peak irradiation the distal edge of the pristine Bragg peak is very sharp. We are therefore designing a ridge filter to reduce the number of energy layers required.

Conclusion The simulations have shown the beam line can deliver a wide range of irradiation modalities. Lateral and distal dose fall-off are commensurate with the small targets. This new research infrastructure will provide state of the art small animal irradiations to the international research community from the second half of 2024.