Application programming interface guided QA plan generation and analysis automation

Abstract Purpose Linear accelerator quality assurance (QA) in radiation therapy is a time consuming but fundamental part of ensuring the performance characteristics of radiation delivering machines. The goal of this work is to develop an automated and standardized QA plan generation and analysis system in the Oncology Information System (OIS) to streamline the QA process. Methods Automating the QA process includes two software components: the AutoQA Builder to generate daily, monthly, quarterly, and miscellaneous periodic linear accelerator QA plans within the Treatment Planning System (TPS) and the AutoQA Analysis to analyze images collected on the Electronic Portal Imaging Device (EPID) allowing for a rapid analysis of the acquired QA images. To verify the results of the automated QA analysis, results were compared to the current standard for QA assessment for the jaw junction, light‐radiation coincidence, picket fence, and volumetric modulated arc therapy (VMAT) QA plans across three linacs and over a 6‐month period. Results The AutoQA Builder application has been utilized clinically 322 times to create QA patients, construct phantom images, and deploy common periodic QA tests across multiple institutions, linear accelerators, and physicists. Comparing the AutoQA Analysis results with our current institutional QA standard the mean difference of the ratio of intensity values within the field‐matched junction and ball‐bearing position detection was 0.012 ± 0.053 (P = 0.159) and is 0.011 ± 0.224 mm (P = 0.355), respectively. Analysis of VMAT QA plans resulted in a maximum percentage difference of 0.3%. Conclusion The automated creation and analysis of quality assurance plans using multiple APIs can be of immediate benefit to linear accelerator quality assurance efficiency and standardization. QA plan creation can be done without following tedious procedures through API assistance, and analysis can be performed inside of the clinical OIS in an automated fashion.


| INTRODUCTION
Quality Assurance (QA) of medical linear accelerators (linac) is a critical responsibility of the qualified medical physicist. The increasing complexity in linac technology warrants standardization in clinically practicable QA. Recommendations to reduce redundancy in testing the parameters characterizing the linac and advise on the frequency at which tests are required have been put forth by many publications, including the Task Group 142 from the American Association of Physicists in Medicine. 1 Although AAPM TG-142 recommendations are available since 2009, the machine QA tests are not implemented consistently even within the same institutions. 2 Efficiency, standardization, and error reduction are a central theme to the development of new QA processes, and the technology has been assisted in developing solutions for safer and more effective QA. 3,4 These solutions lead to effortless generation of QA plans and more efficient and consistent delivery and analysis of these plans. analysis tools outside of the linac software environment. [5][6][7][8] The generation of complex QA plans, such as QA plans for Volumetric Modulated Arc Therapy (VMAT), have been subsidized by vendors and researchers. 9,10 Plans may also be generated and delivered through the clinical Oncology Information System (OIS). Vendor provided application programming interface (API) features allow for the programmatic generation of QA patients and plans. 11,12 A linac's electronic portal imaging device (EPID) demonstrates increased efficiency in high resolution QA measurement by facilitating the acquisition of dosimetric images for patient specific dynamic multileaf collimator (DMLC) QA. 9,13 Output consistency, dose linearity, and high resolution make the EPID effective in image collection for periodic dosimetry and imaging QA and acceptance testing. 3,14 Due to the digital image output from the EPID, automated analysis can be achieved with commercial software applications 15,16 or inhouse software solutions. 17,18 EPIDs have become the standard acquisition devices for common quality assurance data collection.
Through a combination of API technology, EPID data collection, and custom image analysis software, this study demonstrates an automated quality assurance plan creation and evaluation process that can assist clinical institutions in periodic QA. Clinical TPS write-enabled scripting is an efficient way in generating plans, especially QA plans of predefined collimation jaw and MLC positions. Delivering plans through the clinical OIS allows for images acquired during plan delivery to be automatically transferred to the clinical TPS database, thus facilitating an efficient means for analysis with API assisted image analysis software.

2.A | Plan creation
The periodic generation of daily QA plans within our institution's clinical practice is necessitated by a high volume of patient images, requiring the generation of a new QA patient. When the number of patient images within the OIS reaches a maximum capacity, stress on the database leads to stalled image transfer or even software crashes. The plans could also be delivered through a nonclinical mode (no record-and-verify), but the collection and organization of EPID images would be the responsibility of the clinician delivering the QA plan as opposed to the automated transfer and storage of the OIS. A required change in treatment delivery parameters (i.e. treatment machine, energy, or MLC type change) may also require the creation of periodic QA plans. The manual plan creation for daily QA requires several procedural steps (Fig. 1). The time required to complete the daily QA plan creation varies greatly with the experience of the user. The current institutional procedures require monthly and quarterly QA to be delivered through external DICOM files outside of the OIS; any modification required to these DICOM files requires DICOM tag manipulation from custom code.  library, specifically the SimpleShapeChecker class that allows for an object within the image-specified by a contrast boundary with its surroundings-to be fit into an assumed shape (a circle). 31 The picket Junction ¼ I junc À I BG I rad À I BG À 1, where I junc , I BG , and I rad are the intensity values at the junction between the fields, the background reading, and the reading inside

| RESULTS
The AutoQA Builder has been executed a total of 322 times in the 14 months since its deployment (Fig 6).   The tolerances for VMAT QA are for each normalized band to be within AE1.5% of the average of all normalized bands. Table 2 shows the calculated results-the average of each corrected band normalized to the average corrected reading-for the T2 and T3 VMAT tests.
To test the sensitivity of each QA analysis application to intentional errors in the delivery of the quality assurance plans, images were acquired with improper geometry and/or phantom placement.
The junction test was compared against the DoseLab manually calculated junction intensity values. The contingency table below shows that all tests that passed the DoseLab analysis also passed the PDSAPI analysis and is similar for all failing tests (Fig 10). The lightradiation coincidence test had some results (7) that failed in the PDSAPI application, but not in the manual BB selection analysis within DoseLab. This yields a sensitivity of 80.5%.
The impact of intentional errors to the picket fence analysis can be seen in Table 3.

| DISCUSSION
Once a QA plan has been designed, the automated generation of those plans assist in the efficiency and consistency with which the tests can be implemented. This leads to each linac within an institu-  The histogram mean values within each band of the VMAT delivery is expected to be identical between the manual and automated analysis, but errors in user input, determining the correct histogram size, and mathematical rounding may lead to small discrepancies as seen in Table 2

| CONCLUSION
The automated creation and analysis of quality assurance plans using multiple APIs can make an immediate and substantial impact on linear accelerator quality assurance efficiency and standardization.
Keeping the QA plans and images in the same OIS from beam generation, through delivery, and to analysis assists in the organization of acquired images for the physicists performing the analysis and for future auditing of quality assurance. With hundreds of QA plans generated and dozens of monthly and quarterly QA plans analyzed with the assistance of APIs at our institution, the QA delivery and analysis landscape is changing to a more streamlined and homogeneous approach.

AUTHOR CONTRI BUTION
MCS, NCK, BS, and FJR contributed to the initial ideas that lead to the creation and implementation of the methods described in this project. CAM, YW, and MCS contributed to the software F I G . 1 0 . Contingency table between DoseLab manual analysis and portal dosimetry scripting application programming interface for junction tests with intentional gaps and overlaps placed in the field and light-radiation coincidence tests with varying SSD setup.

D A T A A V A I L A B I L I T Y S T A T E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.