A Sparse Orthogonal Collimators System for Experiments on Small-Animal Scale

Abstract

<h3>Purpose/Objective(s)</h3> Intensity-modulated radiotherapy (IMRT) is the clinical gold standard for concentrated dose delivery. However, IMRT for small animal experiments has faced paramount technical challenges, preventing effective clinical translation of preclinical experiments. Sparse orthogonal collimator (SOC) based on the Rectangular Aperture Optimization (RAO) framework is the first general-purpose small animal IMRT platform, but its original planning algorithm was inefficient for complex cases such as total marrow irradiation (TMI) of mice. Here we introduce a significantly improved optimization framework for efficient SOC-based treatment planning and demonstrate TMI efficacy. <h3>Materials/Methods</h3> The SOC system is designed and fabricated with four orthogonal, double-focused tungsten leaf pairs that are capable of forming two rectangles simultaneously. The rectangles were optimized using RAO. Since the Kronecker product representation of the rectangular aperture is the main limitation of the computational performance, we reformulated matrix multiplication in the data fidelity term using multiplication with small matrices instead of the Kronecker product of the matrices. The optimization problem can then be efficiently solved using the Fast Iterative Shrinkage-Thresholding Algorithm (FISTA). Five mice CTs with manually delineated bone marrow and OARs, including the kidneys, lungs, heart, spleen, liver and bowel, were used in the planning study. 12 Gy was prescribed to the bone marrow target while expecting the bones to receive 2.5X higher doses due to kV photoelectric interactions. RAO plans were compared with manual 3D plans with custom lead blocks. <h3>Results</h3> SOC plans markedly improved TMI PTV hot spots and significantly reduced OAR doses compared with other experimental data. The PTV D5, indicating the maximum dose, was reduced by 21.6%. The average doses to the kidneys, lungs, heart, liver, and bowel were reduced by 98.4%, 95.7%, 79.1%, 97.7%, and 82.8%. The average plan optimization total time was 40 minutes. <h3>Conclusion</h3> This study shows the feasibility of creating SOC-based RAO plans for complex and large targets such as TMI. The marked improvement in OAR sparing allows mouse experiments to more closely simulate human treatments, which is critical for TMI, whose treatment response and outcome heavily depend on both the target and off-target doses and toxicity.