Access Type

Open Access Thesis

Date of Award

January 2019

Degree Type


Degree Name



Medical Physics

First Advisor

Di Yan

Second Advisor

Michael Snyder


Purpose: Monte Carlo-based log file quality assurance (LF-MC QA) is investigated as an alternative method to phantom-based patient-specific quality assurance in radiotherapy (e.g. ArcCHECK QA (AC QA)).

Methods: First, the shortcomings of AC QA were investigated. The sensitivity dependence of ArcCHECK diodes on dose rate (in-field) and energy (primarily out-of-field) was quantified. LF-MC QA was then analyzed on the phantom geometry. Planned (‘Plan’) and LF-reconstructed CS and MC doses were compared with each other and AC measurement via statistical (mean ± StdDev(σ)) and gamma analyses to isolate dosimetric uncertainties and quantify the relative accuracies of AC QA and LF-MC QA. LF-MC QA was then analyzed on the patient geometry. Calculation algorithm dependent (Plan-MC vs Plan-CS) and delivery error (LF-MC vs Plan-MC) dependent dosimetric discrepancies were isolated. Dose discrepancies were evaluated using PTV Dmean, D99, and D1 as well as tumor control probability (TCP). Dose discrepancy due to calculation algorithm was further assessed as a function of heterogeneity and beam modulation complexity (MU/Rx). LF QA results were compared to clinical AC QA results. Various LF-MC QA pass/fail protocols were assessed.

Results: Calculation and ArcCHECK measurement differed up to 1.5% in-field due to variations in dose rate and up to 5% out-of-field due to energy effects. On the ArcCHECK geometry, phantom-dependent, calculation algorithm-dependent (MC vs. CS), and delivery error-dependent dose uncertainties were 0.8±1.2%, 0.2±1.1%, and 0.1±0.9% respectively. On the patient anatomy, percent differences in [PTV Dmean, D99, D1] were [-0.1±0.1%, 0.0±0.2%, -0.2±0.2%] for machine delivery error, [-3.4±1.9%, -4.6±2.8%, -1.2±2.8%] for dose calculation difference, and [0.5±2.0%, 0.2±1.2%, 2.6±4.1%] due to limited VMAT beam sampling. Drop in TCP due to calculation difference (MC-CS) was -3.1±1.8% [min -5.7%]. 41% of PTV D99 dose calculation difference was due to beam complexity. Heterogeneity effects were negligible for H&N. For lung, 18% of dose calculation difference on PTV Dmean was due to heterogeneity

Conclusions: ArcCHECK QA was consistently incapable of catching clinically relevant dose discrepancies as calculated on the patient anatomy using LF-MC QA.