Lattice radiation therapy for canine sinonasal carcinoma: feasibility study of spatially fractionated radiotherapy using volumetric modulated arc therapy.

BACKGROUND: Lattice radiotherapy (LRT) administers spatially fractionated doses to a tumor volume with the intent to induce beneficial anti-tumor host responses and avoid normal tissue damage. This study investigated feasibility of LRT for canine sinonasal carcinoma (SC) using volumetric modulated arc therapy (VMAT), aiming to achieve the following criteria: (1) place ≥ 3 vertices of high dose "peak" regions within the gross target volume (GTV) separated by low dose "valley" regions; (2) achieve a 3:1 and 5:1 peak-to-valley dose ratios (PVDR); (3) meet organ at risk (OAR) dose constraints; (4) meet quality assurance (QA) standards. A secondary aim was to assess the impact of dose calculation algorithms and grid size on dosimetry. The peak vertices were placed manually as 5 mm 3D spheres within the GTV. Treatment plans were initially optimized and calculated on Eclipse using Acuros with 2.5 mm grid size and underwent QA using Varian's Portal Dosimetry (3 mm/3%). Plans were then recalculated without re-optimization comparing Acuros XB (AXB) and anisotropic analytical algorithm (AAA) with variable grid sizes, to compare doses estimated for peak and valley lattice target volume and OARs.

RESULTS: Twenty plans were generated from 10 image sets from dogs with SC. In all cases, (1) placement of ≥ 3 vertices was possible, with a median center-to-center distance of 30 mm (range: 26-41 mm); (2) achieved both 3:1 and 5:1 PDVR; (3) plans met OAR constraints; (4) all plans passed QA with gamma index > 95%. When comparing calculation algorithms with a consistent grid size, AAA overestimated the dose in both peak and valley regions compared to AXB. When using AXB, increasing the dose calculation grid size from 1.0 mm to 2.0 mm and 2.5 mm led to underestimation of the lattice metrics and overestimation of the valley dose, ultimately resulting in a reduced calculated PVDR. These deviations were not detected on standard clinical QA with portal dosimetry.

CONCLUSION: VMAT LRT planning and delivery is feasible for canine SC. This data suggests that the predicted-and consequently, reported-lattice target metrics can vary depending on specific treatment planning software settings. Further investigations could define the most appropriate QA approach for high-resolution LRT plans.