An open-source irradiation and data-handling framework for pre-clinical ion-beam research.

Abstract

Context: Pre-clinical animal studies are pivotal for understanding the radiation effects in particle therapy. However, small animal research often relies on highly customized in-house solutions. This study introduces a comprehensive, open-source data processing pipeline specifically developed for pre-clinical particle irradiation research in a multi-vendor setting.

Materials and methods: The pipeline for pre-treatment, positioning, and response imaging encompassed ultra-high-field micro magnetic resonance imaging (Bruker UHF-µMRI), micro-computed tomography (Molecubes µCT), micro-positron emission tomography (Molecubes µPET), and in-room cone-beam computed tomography (MedPhoton CBCT). For image postprocessing, 3D Slicer was primarily used for registration, resampling, cropping, and data conversion from DICOM to the NIFTI format. The RaySearch treatment planning and delivery infrastructure was employed for target delineation and kV x-ray dose calculation (μ-RayStation 8B) and for ion beam treatment planning (RayStation 2024A-SP1-R), as well as in-room image registration and ion-beam delivery (RayCommand 2024A-SPC2). The oncology information system (OIS) RayCare 2024A-SP4 served as the interface between the treatment control system and treatment-planning-system (TPS). DICOM metadata from commercial products had to be modified to allow import and processing for this customized pre-clinical workflow. Anatomical mouse atlases were integrated to support post-irradiation analysis using structure sets, dose and dose-averaged linear energy transfer (LET_D) maps were also processed. Five mice, irradiated unilaterally with a single 60 Gy proton dose, were used for workflow validation. Radiomic features were explored to assess irradiated and non-irradiated hippocampal tissue.

Results: A python-based DICOM processor successfully converted vendor-specific 3D multi-frame objects into a TPS compatible format while preserving the global unique-identifier (UID) hierarchy and spatial metadata. This enabled a seamless image data import into the TPS. All multimodality images, planning-related structures, doses, and LET_D maps were stored on an internal data repository after processing in 3D Slicer. The framework was applied to five mice, demonstrating that the contralateral median dose remained ≤ 0.03 Gy, including follow-up µMRI and exemplary radiomics analysis.

Conclusion: The open source in-house developed irradiation and data-handling framework¹ enabled the establishment of a technological platform for image-guided pre-clinical ion-beam research. All steps from treatment planning, position verification, and beam delivery, down to non-invasive response assessment were covered.