Vertex-Wise 3D QA for Lattice Radiotherapy Using Segmented Isodose Structure Analysis
Abstract
Purpose
This study proposes a vertex-wise Isodose Structure (IS)-based framework to evaluate the 3D geometric accuracy of Lattice Radiotherapy (LRT) dose distributions by decoupling geometric misalignment from dose-dependent volumetric distortion, enabling vertex-specific 3D QA endpoints.
Methods
LRT plans were delivered to a cubic phantom containing vGEL (Triangle Products, Japan) using TrueBeam (Varian, USA) and OXRAY (Hitachi, Japan) accelerators. Measured dose distributions were compared with calculated distributions using gamma analysis (3%/2mm/10%) and IS analysis. For IS analysis, isodose volumes from 10% to 120% were reconstructed. Volume Ratio (VR, the ratio of the calculated isodose volume to the measured volume) and centroid displacement of isodose structures were evaluated using (1) Global Vertex Analysis (GVA), treating the full distribution as a single structure, and (2) Local Vertex Analysis (LVA), isolating individual vertices, referenced to planned vertex centroids. Vertex boundaries were segmented by connecting the local minima between neighboring vertices.
Results
Gamma passing rates (GPR) were 93.4% (TrueBeam) and 96.4% (OXRAY). A total of 27 vertices were analyzed per delivery across 12 isodose levels (10–120%) for each system. Despite high GPR, IS analysis revealed notable discrepancies. VR exhibited strong dose dependence, decreasing from ~1.0 at low doses to ~0.77 (TrueBeam) and ~0.67 (OXRAY) at 100%, indicating systematic, non-uniform expansion in high-dose regions. This dose-dependent volumetric expansion biased the global centroid, producing apparent global centroid displacements that may not reflect vertex-wise misalignment. Accordingly, GVA showed global centroid displacements of 2.6mm (TrueBeam) and 2.1mm (OXRAY), whereas LVA yielded median vertex centroid displacements of <1mm.
Conclusion
Gamma analysis may fail to detect geometric errors in LRT QA. GVA can similarly misinterpret non-uniform dosimetric distortions as geometric setup errors. In contrast, LVA effectively decouples true geometric accuracy from local dosimetric variation, providing a more reliable 3D verification framework and enabling actionable vertex-wise QA endpoints.