The sensitivity and specificity of patient-specific QC at the Wellington Blood and Cancer Centre.

Type of content
Theses / Dissertations
Publisher's DOI/URI
Thesis discipline
Medical Physics
Degree name
Master of Science
Publisher
University of Canterbury
Journal Title
Journal ISSN
Volume Title
Language
English
Date
2017
Authors
Scarlet, Benjamin Ewen William
Abstract

Purpose Patient-specific quality control (QC) plays an important role in assuring the safety of treatment planning and delivery for complex treatment techniques such as volumetric modulated arc therapy (VMAT). Ideally, patient-specific QC should be able to detect clinically relevant errors in treatment plans (sensitivity), pass treatment plans that do not contain errors (specificity), and resolve different error modes. Previous studies in literature have reported that patient-specific QC methods have a lower sensitivity than specificity. Therefore, the aim of this study was to quantify the sensitivity and specificity of the patient-specific QC methods currently available at the Wellington Blood and Cancer Centre. A secondary aim was to determine if any QC method could resolve different error modes.

Methods Intentional errors simulating incorrect linac monitor unit delivery (MU), multi-leaf collimator (MLC) positioning, dosimetric leaf gap (DLG), focal spot size (FSS) and output variation with gantry angle were introduced to five Head & Neck VMAT plans. Criteria were defined to determine whether each introduced error caused a clinically relevant dose deviation to the patient treatment. Error-free plans and introduced error plans were measured using a time resolved point dose method (trPD), an EBT3 film method (Ashland Inc.), and using an ArcCheck helical array (Sun Nuclear corp.). Sensitivity of each QC test (true positive rate) and intrinsic measurement system sensitivity (change in output over change in input) were calculated, as well as the specificity of each QC test (true negative rate). In addition, receiver operator characteristic (ROC) curves were created for each QC method, and the efficiency of each QC method was determined by calculating the area under the ROC curve (AUC). In addition, the ability to resolve different error modes was investigated for the trPD method.

Results A total of 89 plans were created (5 error-free, 84 containing introduced errors). Of the 84 introduced errors, 52 caused clinically relevant dose deviations. Using the clinically applied QC acceptance criteria (±2% dose difference for trPD, >85% of points passing a {2%;2mm} -criterion for film and ArcCheck), all three QC methods were found to have high specificity but the sensitivity was comparatively low (Table 1, first row).By varying the acceptance criterion for trPD to ±1.9%, and modifying the beam model and acceptance criterion to 87% for film, the sensitivity of these methods could be improved at the expense of a slight decrease in the specificity. However, the observed improvements in efficiency were within the estimated uncertainty range. For the ArcCheck system, none of the investigated configurations yielded an acceptable sensitivity and specificity (Table 1, second row, optimal practise involved varying the passing criterion to 92%). Investigation of the ArcCheck intrinsic sensitivity indicated that these poor results were caused by a systematic offset between the measured and TPS calculated dose. Analysis of the trPD results showed that different error modes could potentially be resolved by using a per region analysis, where the detector distance to the MLC field edge defined the different regions.

Conclusion The current patient-specific QC methods at the WBCC displayed a low sensitivity for clinically relevant errors but a high specificity, similar as reported by other published studies. This study showed the importance of quantifying the characteristics of patient-specific QC methods in more detail prior to clinical application.

Description
Citation
Keywords
Ngā upoko tukutuku/Māori subject headings
ANZSRC fields of research
Rights
All Rights Reserved