Program Information
The Application of Statistical Process Control and Incident Learning to Monthly Linac CBCT Image Quality QA
R Manger*, J Hoisak , G Kim , University of California, San Diego, La Jolla, CA
Presentations
TH-AB-FS1-15 (Thursday, August 3, 2017) 7:30 AM - 9:30 AM Room: Four Seasons 1
Purpose: The AAPM TG-142 specifies cone-beam computed tomography (CBCT) image quality (IQ) metrics to be monitored at various frequencies using tolerances derived from baselines obtained at acceptance/commissioning. This study focuses on monthly CBCT IQ metrics and their tolerances. The purpose of this study is to develop custom IQ tolerances for CBCT, compare them with acceptance/baseline tolerances and determine if either action threshold correlates with clinically-reported IQ issues.
Methods: Several years of IQ data were acquired for eight linac-based CBCT systems using the Catphan 500 with standardized acquisition parameters, and analyzed with the TotalQA software. Statistical process control (SPC) was applied to a stable six month period of data to develop action thresholds for Geometric Distortion, Spatial Resolution, HU Constancy, Uniformity, and Noise. Our institution’s incident learning system (ILS) was also queried to identify clinically relevant CBCT IQ events and preventative maintenance and service. These events were compared to monthly CBCT IQ and SPC-derived action thresholds to determine if there was any correlation.
Results: Across 8 machines, the mean SPC-derived action threshold ranges for Geometric Distortion, Spatial Resolution, HU Constancy, Uniformity, and Noise were 0.16-0.424, 4.75-7.39, 5.66-19.71, 3.30-23.78 and 7.46-13.77, respectively. These threshold ranges were considerably tighter than the acceptance/baseline-derived action levels. Our ILS logged 14 events related to CBCT IQ. For all but one event, CBCT IQ monthly QA results could not reveal the image problem with the acceptance/baseline action levels.
Conclusion: SPC may be used to set machine-specific action thresholds for monthly imaging IQ tests that are more stringent than acceptance/baseline values. However, clinically relevant CBCT issues were not detected by either SPC- or acceptance/baseline-derived action levels applied to monthly IQ results. Further study is warranted to develop CBCT IQ tests that can detect deviations before they become clinically relevant.
Contact Email: