Beam model development and clinical experience with RadCalc for treatment plan quality assurance in online adaptive workflow with an MR‐linac

TL;DR

This paper describes optimizing RadCalc software for accurate dose verification in MR-linac treatments, improving clinical workflow efficiency and safety.

Contribution

The paper introduces a validated beam model for RadCalc in an MR-linac workflow, enabling reliable automated dose verification.

Findings

RadCalc calculations agreed with measurements within 2% for simple geometries and 5% for complex geometries.

Validation against the Monaco TPS showed a mean point dose difference of −0.3 ± 0.8% across 39 clinical plans.

Retrospective analysis of 4085 plans showed 0.3% fail rate for individual fields with discrepancies >10%.

Abstract

The aim of this work was to report on the optimization, commissioning, and validation of a beam model using a commercial independent dose verification software RadCalc version 7.2 (Lifeline Software Inc, Tyler, TX, USA), along with 4 years of experience employing RadCalc for offline and online monitor unit (MU) verification on the Elekta Unity MR‐linac (MRL) for a range of clinical sites.

Calculation settings and model parameters, including the Clarkson integration settings and radiation/light field offset, have been systematically examined and optimized, and pitfalls in the use of density inhomogeneity corrections and in off‐axis calculations were investigated and addressed. The resulting model was commissioned by comparing RadCalc calculations to measurements for a variety of cases, selected following relevant recommendations, ranging from simple fields in a water tank to end‐to‐end point dose measurements in an anthropomorphic phantom.

For simple geometries, the agreement was within 2%, and for complex geometries, within 5%. When validating against the Monaco (Elekta AB, Stockholm, Sweden) treatment planning system (TPS), for 39 clinical commissioning plans, the mean total point dose difference was −0.3 ± 0.8% (−2.0%–1.1%). Finally, when applied retrospectively to 4085 clinical plan calculations, the agreement with the TPS was 0.3 ± 1.1% (−4.8%–4.2%), with fail rates of 0.1% for total point dose (discrepancy > 4%) and 0.3% for individual fields (discrepancy > 10%).

Improved calculation agreement with the TPS and therefore increased confidence in the online QA, opened the way for an automated and physics‐light independent MU verification workflow within our MRL program.

Figures