The IROC Houston anthropomorphic phantom program is an end-to-end test of the radiotherapy process. An institution receives the phantom and treats it like a patient, including simulation, planning, and delivery of a treatment. The head and neck phantom, in particular, is a staple for verification of IMRT dose delivery. This phantom has been irradiated over 2,000 times to credential institutions wishing to participate in IMRT-based clinical trials as well as for institutions simply wishing to verify the accuracy and integrity of their IMRT program.
While the pass rate for this phantom has increased over the years,1 10% of institutions still fail to deliver the dose they believe they are delivering. And this failure rate is based on a relatively lax 7% point dose agreement criteria in the target and planar criteria of >85% of pixels passing a 7%/4mm gamma analysis. The failure rate increases to 30% if the criteria are tightened to 5%/3mm.2
The end-to-end nature of the phantom irradiation process is both a strength and a weakness. While it tests the entire process, by the very fact that all parts of the process are involved, it is very hard to identify the underlying cause of any given error that is observed. Why someone fails the phantom, therefore, is a particularly challenging question to answer. However, the general findings on this question are concerning. Qualitatively, the vast majority of phantom irradiations that fail to meet criteria reveal that the delivered dose distribution is of the right shape, and in the right location, it just has the wrong magnitude.2 Systematically, too much or too little dose is delivered compared to what has been planned. This is concerning because it is hard to imagine that such an error would not also manifest in patient cases.
At the IROC Houston QA office we are developing infrastructure to help us understand the causes of specific phantom failures. Ultimately, we expect this to help us identify problems and work with institutions to remedy them, thereby providing optimal and consistent radiotherapy for all patients. As a first step, we have developed an independent recalculation tool for phantom plans. Using the Mobius 3D calculation platform, we have created reference models for each make and manufacturer of linear accelerator based on reference measurements on thousands of linacs during IROC Houston site visits.3 We can then use this platform to recalculate phantom plans. Of particular interest are cases where the institution's dose calculation agrees poorly with the dose that was actually delivered, and significantly better agreement is seen in the dose distribution from our recalculation. Such a situation is indicative of a dose calculation error by the institution. Initial testing of this approach on 259 recent head and neck phantom cases found that in 45 cases (17%), the recalculation was significantly and substantially more accurate (>2% average improvement) than the institution's TPS calculation. Of the 19 phantoms that failed IROC Houston's criteria, our independent recalculation was significantly and substantially more accurate for 13 of them (68%). It was more accurate by more than 3% on average over these 19 phantoms, including one case where the recalculation was 6.5% more accurate.4
These results are concerning in terms of IMRT dose calculation accuracy, and highlight that IROC needs to carefully examine aspects of dose calculation (such as beam modeling) when examining a failure of the head and neck phantom. Ongoing work to dig further into this issue and understand the actual underlying issues includes doing a sensitivity analysis of TPS parameters on performance of the head and neck phantom, and statistical analysis of the root causes of failure. Hopefully, in the not-too-distant future, in the case of a failing result from a phantom irradiation we will be able to isolate the specific probable cause(s) and thereby provide direct and helpful feedback to the institution.
We have noticed that you have an ad blocker enabled which restricts ads served on this site.
Please disable it to continue reading AAPM Newsletter