Encrypted login | home

Program Information

Automated and Streamlined KV CBCT Image QA: A Multi-Institutional Study

A Ayan

A Ayan1*, M Whitaker2 , H Al-Hallaq3 , S Hsu4 , J Woollard5 , D Roberts6 , N Shtraus7 , N Gupta8 , J Moran9 , G Kim10 , (1) Ohio State Univ, Columbus, OH, (2) Image Owl, Inc., Greenwich, NY, (3) The University of Chicago, Chicago, IL, (4) Montefiore Medical Center, Bronx, NY, (5) Arthur G. James Cancer Hospital, Hilliard, OH, (6) University of Michigan Hospital, Ann Arbor, MI, (7) Tel Aviv Medical Center, Mazkeret Batia, ,(8) Ohio State Univ, Columbus, OH, (9) Univ Michigan Medical Center, Ann Arbor, MI, (10) University of California, San Diego, La Jolla, CA


TH-AB-FS1-4 (Thursday, August 3, 2017) 7:30 AM - 9:30 AM Room: Four Seasons 1

Purpose: To develop and test an infrastructure to support standardized and automated testing and analysis of CBCT image quality QA in a framework supporting multiple institutions.

Methods: QA of linac CBCT imaging systems is an important part of an overall QA program with recommendations provided by AAPM Task Group reports 142 and 179. The automated software platform was developed by a vendor with input and validation from 6 institutions participating in a consortium dedicated to automated QA. The image upload and analysis processes are automated using a MATLAB-based platform which eliminates any dependence on individual users. The platform monitors a user-defined network location for the existence of new CBCT images. When new images are found, they are automatically sorted, categorized in the case of multiple series and then uploaded to a cloud based central database via a REST API using HTTP requests. When image analysis is completed, a standardized analysis report is generated and automatically downloaded back into user’s defined location for further evaluation and analysis.

Results: The developed platform has been piloted by 6 institutions (16 TrueBeam machines) and has demonstrated the ability to perform standard CBCT image QA analysis across multiple institutions. Results can be viewed either by institution or in comparison with all linacs in the consortium’s database. Data are plotted as a function of means with the upper and lower control limits defined per scan protocol. For example, quality metrics of geometric distortion, spatial resolution, contrast, HU constancy, uniformity and noise have been calculated 0.14±0.27mm, 4.98±1.11lp/cm, 3.57±4.52mm, 13.5±11.4HU, 15.0±25.0HU, 12.0±12.2HU, respectively.

Conclusion: Through this platform, outlier results can be immediately identified over a larger cohort of linear accelerators. This work will be used to assess the best frequency of tests as well as the sensitivity of different tests in support of a broader QA program.

Contact Email: