Answers are grounded in this paper and cite their sources.
Objective: The aim of this study is to establish an automatic delineation model of clinical target volume (CTV) in postoperative cervical cancer using data training algorithm to improve the efficiency and the quality of CTV contouring among clinicians with different skilled levels in various radiotherapy institutes. And to improve the existing automatic pelvic bone (PB) delineation. Methods: A total of 145 postoperative patients with early stage (IB1-IIA2) cervical cancer from 4 different hospitals, which are the Army Medical Center of PLA (Da Ping hospital) and other three integrated hospitals, were enrolled in this study. The CTV and the PB were contoured by experienced radiation oncologist. Of these 145 cases, 30 from our medical center were used for autonomous training and the remaining 115 for model testing. Three evaluation indicators, including the mean Dice similarity coefficient (DSC), Hausdorff distance (HD) and relative absolute volume difference (RAVD) were used to evaluate the accuracy of automatic delineation model. Subjective evaluation was also involved in the improvement of PB automatic delineation model. Results: The mean DSC value of our center was 0.861 for the delineated CTV. The mean 95HDand RAVD values of CTV were 10.677 mm and 6.811%. The external verification for the delineated CTV indicated that CTV structure were well delineated in three integrated hospitals. The mean DSC values were 0.773, 0.795 and 0.738 respectively. Tested with the existing automatic PB delineation model, the new PB contouring model has a relatively low but consistent mean DSC value, which is 0.733. The clinical professionals’ subjective assessments considered that the new PB contours model was acceptable for clinical application. Conclusion: Our deep learning (DL)-based automatic delineation models of CTV and PB in postoperative early staged cervical cancer are efficient and accurate for clinical usage.
No takes yet. Share an insight, caveat, or question.
Xu et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: