Characterizing the Spatial Distribution of Surface Dose In Head and Neck Radiotherapy
Abstract
Purpose
In head-and-neck (H&N) cancer radiotherapy, using IMRT and VMAT, treatment evaluation relies primarily on dose–volume histograms DVH. While effective for quantifying dose magnitude, DVH discard spatial information that may be critical for understanding toxicity such as acute radiation dermatitis. This study focuses on characterizing the spatial organization of surface dose, including heterogeneity, fragmentation, and local gradient structure.
Methods
For 372 head-and-neck cancer patients treated with IMRT/VMAT at MUHC, dose was sampled within a 0–3.0 mm subsurface shell and mapped onto a three-dimensional skin surface mesh, from which dose surface histograms (DSHs) were generated to derive surface mean dose Smean, and Sx , defined as the percentage of skin receiving at least X Gy. To capture spatial characteristics of surface dose, surface dose gradients were computed from local dose differences along the surface mesh. For higher-order spatial analysis, the surface dose was unwrapped into a two-dimensional dose surface map (DSM) using a cylindrical parameterization aligned with the longitudinal axis of the neck. From these DSMs, dosiomic features were extracted, including Gray-Level Co-occurrence Matrix (GLCM) entropy and energy, Gray-Level Size Zone Matrix (GLSZM) number of zones, and Gray-Level Run Length Matrix (GLRLM) features.
Results
Skin surface dose exhibited multimodal exposure regimes across the cohort as reflected by Smean. Within individual Smean modes, patients demonstrated different spatial organizations of dose on the 3D skin surface and differences in local dose transition quantified by the mean edge gradient. DSM-based texture metrics, particularly GLCM energy, captured this variability by measuring surface dose uniformity. Patients with low GLCM energy exhibited fragmented high-dose regions, whereas patients with higher GLCM energy showed large, spatially contiguous dose regions.
Conclusion
Surface-based dose representations enable characterization of the spatial organization and heterogeneity of skin dose, offering potential to improve skin toxicity risk assessment in H&N radiotherapy.