Author ORCID Identifier
Date of Graduation
Doctor of Philosophy (PhD)
Stephen F. Kry PhD
Peter A. Balter PhD
Rebecca M. Howell PhD
Christine B. Peterson PhD
Julianne M. Pollard-Larkin PhD
The Imaging and Radiation Oncology Core (IROC) phantoms are used as an end-to-end test of an institution’s radiotherapy processes, and for clinical trial credentialing. Phantoms are treated like patients, and evaluation of the doses received by the thermoluminescent dosimeters (TLDs) inside the phantom, reflects the accuracy with which an institution can image, plan and irradiate a phantom or patient. Recent phantom results show that among the hundreds of various IROC phantoms irradiated annually, 8-17% of institutions fail this test. The purpose of this work was to investigate the various types of errors that may occur during the treatment process and quantify the magnitude of their contribution to planned treatment planning system (TPS) to measured TLD phantom dose deviation (TPS vs TLD dose deviation).
First, a preliminary study was conducted to identify the causes of failures among IROC phantoms. Categories of failure were established, and phantoms grouped accordingly. The results of this study lead to the investigation of three major error contributors: dose calculation error, delivery error and machine output error. Dose calculation error was assessed through independent recalculation of the phantom plans using a dose recalculation system (DRS). An acrylic output block containing TLDs was irradiated by each institution prior to phantom irradiation, to measure machine output on that day. Machine output error was determined through an assessment of both the output block’s measured TLD doses and the machine output dose reported by the institution using their in-house QA device or ion chamber. Delivery error was assessed by evaluating the machine log files associated with a plan delivery. Prior to collecting log files from institutions, a study was conducted to test the ability of the IROC phantoms to accurately capture log file (specifically MLC) errors. This study used the deliberate introduction of MLC errors into the plans, to assess how well they would translate to measured and log file dose deviations. Delivery log files from institutions irradiating the phantoms were collected and evaluated for MLC rms error and delivered dose error. All error types were assessed on an individual TLD basis. Results were categorized into two groups: TLDs with dose deviations greater than the threshold for TLD measurement uncertainty (3.2%) represented the poorer performing group of phantom TLDs, and those with dose deviations less than 3.2%, the better performing group of phantom TLDs.
The majority (60%) of spine and head and neck (H&N) phantom failures, which are static (no motion) and generally have more highly modulated plans, were caused by systematic dose errors. This was when the dose in the entire plan was either too high or too low throughout the entire plan, indicating errors in the institution’s TPS dose calculations. The lung phantom, which moves to simulate patient breathing, failed primarily due to localization errors. Localization errors, which manifested as the correct amount of dose, but delivered to a location off-set from the PTV, represented 62% of lung phantom failures. Dose calculation errors were found in 47% of all spine phantom results and 42% of all lung results. However, among failing phantoms, this error was present in 93% of spine cases and only 35% of lung cases, indicating a greater impact of dose calculation error on the highly modulated spine treatment versus the lung. Machine output error showed positive correlations with increasing dose deviation for spine (r = 0.55, p
Overall, dose calculation error was found to be the greatest contributor of dose deviations among highly modulated static phantom irradiations (spine and H&N), output error contributed almost equally to all three phantoms and delivery error was minimal with no correlation to phantom performance. Lung phantoms are primarily plagued by motion management related dose deviations which are more difficult to quantify and assess via a remote phantom audit program such as IROC’s. Therefore, among the errors evaluated, which were dosimetric in nature, we were able to quantify 56% of error among H&N phantoms, 68% among spine and only 19% of lung dose deviations.
phantom error, IROC phantom, TLD