Room: Track 3
TG-218 proposes a 3%/2mm threshold for ?-analysis of IMRT QA events and specifies a universal tolerance of 95% and action limit of 90% for ?-pass rate. The report also recommends tracking ?-pass rates for IMRT/VMAT patient-specific QA across patients, looking for systematic errors in the QA regime, and using a statistical control process to determine a local action limit. This work proposes the use of a free, open-source data mining tool to analyze IMRT reports and determine local action levels.
The IMRT QA Data Mining (IQDM) tool uses Python code to parse IMRT QA reports, currently supporting ScandiDos Delta4 and SNC Patient. This software then generates an interactive dashboard including a time-plot with trending, a histogram, and a control chart per TG-218; the charting variable may be ? pass-rates or any other variable parsed from the reports.
Over 10,000 QA reports from six radiation oncology departments were imported and analyzed. TG-218 recommends determining control limits via a small amount (~20) of IMRT measurements while the process is in control. Clinics were able to determine action and control limits for IMRT QA processes. One clinic found for their QA scheme using both 2%/3 mm and 3%/3 mm tolerances as well as global normalization, the action limit was determined to be 90.1% and a lower control limit of 93.7%
Through statistical process control charts, a local control limit can be established for individual clinics per TG-218 recommendations. The IQDM software assisted in identifying systemic biases in the TPS of one clinic as well as deviations in pass/fail criteria from IMRT QA policy.
Quality Assurance, Software, Statistical Analysis