Implementation of an offline adaptive radiotherapy software for head and neck cancer: commissioning and evaluation of dose calculation based on synthetic CT (sCT)

The object of this work is the implementation of an innovative offline adaptive radiotherapy software, ART.1, developed to overcome the inherent limitations of Cone Beam Computed Tomography (CBCT) imaging. To demonstrate its clinical validity, an experimental commissioning process was proposed in alignment with the protocol executed for standard CBCT, aiming to verify the clinical acceptability of contour propagation, dose calculation, and synthetic CT (sCT) generation performed by ART.1. Specific comparisons between the standard radiotherapy Treatment Planning System (TPS) and the ART.1 workflow were executed to evaluate the sCT against the planning CT (pCT) using specific metrics (DSC and SSIM). To achieve this, specialized Python codes sourced from GitHub were first customized and then validated by running them on phantom images to ensure metric calculation accuracy. To test both the responsiveness and robustness of ART.1, different clinical scenarios were investigated by simulating typical patient changes during RT treatment; anatomical variations (e.g., weight gain or loss) were simulated by adding tissue-equivalence bolus of varying thicknesses, and misalignments were replicated through manual shifts. Several tests were performed using different body mask thresholds and slices per image. A statistical analysis of the obtained DSC and SSIM values has demonstrated a general validity of the sCT of ART.1 in correctly managing the presence of bolus, longitudinal shifts, and small rotations. Less satisfactory performances were ob served when a lateral or a vertical shift was present. Here, the results showed values well below the theoretical minimum expected, demonstrating a failure of ART.1 to handle these types of misalignment.