Application analysis of deep learning-based rtStation software in automatic delineation of the heart in patients after surgery for breast cancer(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2021年第6期
- Page:
- 661-665
- Research Field:
- 医学放射物理
- Publishing date:
Info
- Title:
- Application analysis of deep learning-based rtStation software in automatic delineation of the heart in patients after surgery for breast cancer
- Author(s):
- SHI Feiyue1; 2; WANG Min1; ZHAO Ziting3; QIN Wei1; ZHAO Huanyu1; WEI Xiaowei1
- 1. Radiation Therapy Center, Nanjing First Hospital, Nanjing Medical University, Nanjing 210006, China 2. Research Center of Medical Physics, Nanjing Medical University, Nanjing 210029, China 3. Department of Medical Equipment, Nanjing First Hospital, Nanjing Medical University, Nanjing 210006, China
- Keywords:
- Keywords: breast cancer rtStation automatic delineation deep learning heart
- PACS:
- R312；R811
- DOI:
- DOI:10.3969/j.issn.1005-202X.2021.06.001
- Abstract:
- Abstract: Objective To test and evaluate the accuracy of rtStation software based on deep learning for automatic delineation of the heart in patients after surgery for breast cancer. Methods Forty patients after surgery for breast cancer were enrolled in the study, and then divided into 4 groups, namely left breast conserving surgery (LB) group, right breast conserving surgery (RB) group, left radical mastectomy (LG) group and right radical mastectomy (RG) group. After that the CT images for localization were transferred to rtStation software, the heart was automatically delineated by rtStation software for each patient, and the obtained structure files were transferred back to Eclipse treatment planning system. The volume difference (ΔV%), deviation of centroid (DC) and Dice similarity coefficient (DSC) between automatic delineation and manual delineation were calculated for evaluating the accuracy of automatic heart delineation using rtStation software. Results The ΔV%, DC and DSC of the 40 patients were (3.49±10.30)%, (0.48±0.30) cm and 0.89±0.04 respectively. The DSC of the 4 groups were 0.88±0.03 (LB), 0.89±0.03 (RB), 0.88±0.05 (LG) and 0.89±0.04 (RG), respectively and that of 4 combination groups were 0.89±0.03 (LB+RB), 0.88±0.04 (LG+RG), 0.88±0.04 (LB+LG) and 0.89±0.03 (RB+RG), respectively. The statistical analysis results showed that no statistical difference was found in ΔV%, DC and DSC among LB, RB, LG and RG groups (P>0.05), between LB+RB group and LG+RG group (P>0.05), and between LB+LG group and RB+RG group (P>0.05). Conclusion For postoperative patients with different types of breast cancer, using rtStation software to automatically delineate the heart can achieve accurate delineation results, and there is no statistical difference in delineation accuracy among different groups.
Last Update: 2021-06-29