Clinical Validation of Methods for Improving the Quality and Efficiency of Radiation Therapy Contouring Based on RT Contour QA
Abstract
Purpose
This study proposes an automated quality assurance (QA) method for radiation therapy structure delineation based on the RT contour QA software, addressing issues such as low efficiency in delineating clinical radiation therapy regions of interest (ROIs), significant inter-physician variability, and problems with automated delineation like missed detections and misclassifications.
Methods
A structural QA step was integrated into the original radiotherapy workflow, with the RT contour QA software providing six types of abnormality detection rules for radiotherapy structures: bad points, Slice Gaps, overlaps, empty ROIs, ROI cavities, and volume anomalies. This study included the ROI contouring results of 551 radiotherapy patients with six common cancer types, such as nasopharyngeal carcinoma and rectal cancer, over the past two years. The cases were divided into three periods: 2024H1, 2024H2, and 2025H1. Using the software, six types of ROI abnormalities were detected for each period. The average number of abnormality types per period and the incidence rates of different abnormal structures were compared and analyzed to retrospectively evaluate the QA effectiveness of the software.
Results
During the period from 2024H1 to 2025H1, the overall average number of anomaly types showed a gradual decline from 2.25 to 1.79 and then to 1.66, The effect of QA tends to stabilize in the later stage. Among the cancer types, rectal cancer showed the most significant improvement in QA, with the average number of anomalies decreasing by 45.28% year-on-year, then rectal cancer and cervical cancer. The occurrence rate of ROI cavities decreased from 7.10% to 0%, with notable improvements in empty ROIs and bad points.
Conclusion
This study confirms that the automated QA software can quickly identify potential anomalies in delineation, review the lists of automatically delineated ROIs, and provide clear modification suggestions for manually delineated contours, demonstrating significant value in enhancing the precision of radiation therapy.