Author: Sijuan Huang, Zi LIU, Jing MA, Xin Yang 👨🔬
Affiliation: Sun Yat-sen University Cancer Center; State Key Laboratory of Oncology in South China; Collaborative Innovation Center for Cancer Medicine; Guangdong Key Laboratory of Nasopharyngeal Carcinoma Diagnosis and Therapy, Guangzhou, Guangdong, 510060, China 🌍
Purpose: To reduce the pressure of manual review of abnormal situations in manually/automatically delineated regions of interest (ROI), improve its accuracy and basis.
Methods: 220 radiotherapy patients were collected in this study. To improve delineation accuracy and assist physicists in designing treatment plan, based on the RT-Contour QA platform, new rules of quality control (QA) was constructed and continuously improved: (1) isolated areas with a circumference less than 2 millimeters; (2) Discontinuous delineation beyond the specified layer thickness in the scanning protocol; (3) The intersection of anatomical boundaries of organs and tissues without physiological inclusion relationships; (4) Structures that have been named but not delineated; (5) The non closed areas contained within the closed boundary; (6) A mismatch with the preset organ reference volume range. Conducting a structural review of radiotherapy and detecting six common manifestations of ROI can generate structural QA analysis reports. In addition, by analyzing the specific manifestations and causes of delineation errors through statistical analysis, further deficiency in delineation can be identified and be the foundation of subsequent continuous radiotherapy improvement.
Results: The statistical results are as follows: Among all cases, there are 94.67% instances of contouring overlaps, 51.33% occurrences of abnormal volume, 50% instances of structure tomography, 24% occurrences of structural cavities, 7.33% scattered dead pixel, and 5.33% void ROI structures. These results indicate that the primary issues with ROI contouring lie in radiologists' insufficient precision when distinguishing between shapes and boundaries of adjacent organs and tissues; this often leads to omissions or misplacements during manual contouring resulting in suboptimal dots and cavities.
Conclusion: Structure errors during the ROI contouring process, such as abnormal organ volumes and overlapping adjacent organs, directly decrease the quality of plan designed by medical physicists. The RT-Contour QA Platform can help the physicians and medical physicists to check the ROI.