Design and implementation of an automated quality assurance tool for Hounsfield unit-to-relative electron density calibration in cone beam computed tomography imaging.

Abstract

Accurate calibration of Hounsfield Units (HU) to relative electron density (RED) is critical for reliable absorbed dose calculation in radiotherapy, particularly for cone-beam CT (CBCT)-based adaptive workflows. We developed and evaluated a fully automated HU-RED calibration and quality assurance (QA) workflow for Elekta CBCT systems, integrating PyLinac with QATrack + to enable reproducible, user-independent analysis. Using the CIRS 062 M phantom, protocol-specific HU-RED calibration curves were generated for head, thorax, abdomen and pelvis CBCT acquisitions. Systematic protocol-dependent HU offsets were observed relative to fan-beam CT, with air-equivalent inserts shifting from approximately - 1000 HU on CT to + 500-600 HU for body CBCT protocols and up to + 1000 HU for head CBCT acquisitions. Water-equivalent inserts exhibited offsets of + 580-720 HU, while dense bone showed increases of + 200-500 HU depending on protocol. Despite these offsets, short-term reproducibility was high, with standard deviations ≤ 3 HU for water, ≤ 5 HU for bone and ≤ 12 HU for air across repeated scans. Over six months of monitoring, no significant HU drift was observed, demonstrating the ability of the automated workflow to reproducibly measure and track HU constancy over time. Automated tolerance checking using ± 30 HU (water/soft tissue) and ± 50 HU (lung/bone) thresholds showed 100% compliance across all measurements. The complete analysis and QA reporting required less than one minute per scan. This automated approach provides quantitative, protocol-specific HU-RED calibration and stable long-term monitoring, supporting routine CBCT QA and adaptive radiotherapy in accordance with international guidelines.