Correlation between geometric parameters and dosimetric parameters in the evaluation of image auto-segmentation for radiotherapy(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2021年第5期
- Page:
- 540-544
- Research Field:
- 医学放射物理
- Publishing date:
Info
- Title:
- Correlation between geometric parameters and dosimetric parameters in the evaluation of image auto-segmentation for radiotherapy
- Author(s):
- YU Hang; LIU Huan; FU Yuchuan
- Department of Radiotherapy, West China Hospital of Sichuan University, Chengdu 610041, China
- Keywords:
- Keywords: radiotherapy auto-segmentation geometric parameter dosimetric parameter
- PACS:
- R318
- DOI:
- DOI:10.3969/j.issn.1005-202X.2021.05.003
- Abstract:
- Abstract: Objective To investigate and discuss the necessity of combining geometric parameters and dosimetric parameters in the evaluation of image auto-segmentation for radiotherapy by analyzing the relationship between geometric parameters and dosimetric parameters of regions of interest (ROI). Methods An auto-segmentation model established by convolutional neural network was used for the auto-segmentation of organs-at-risk and target areas for 18 patients who received postoperative radiotherapy for cervical cancer, and then the auto-segmentation results were compared with manual segmentation results. The geometric parameters used for evaluation included volume/area-based parameters (Dice similarity coefficient, relative volume difference) and distance-based parameters (maximum Hausdorff distance, 95% Hausdorff distance, centroid difference), while the dosimetric parameters used for evaluation included the average dose difference for organs-at-risk, as well as the ΔD95 and ΔD98 for target area. The relationships between geometric parameters and dosimetric parameters of ROI under different segmentation methods were analyzed by linear regression method and the correlation between geometric parameters and the dosimetric correlation between manual segmentation and automatic segmentation were obtained by Spearman correlation analysis. Results The relationship between geometric parameters and dosimetric parameters of all ROI was weak (63.3% of R2<0.4) and unstable and meanwhile, the correlation coefficient |r| between geometric parameters did not exceed 0.625, indicating weakly correlated or not correlated with each other. Conclusion Both geometric parameters and dosimetric parameters should be concerned when evaluating the results of image segmentation for radiotherapy, and the combination of area/volume-based evaluation method and distance-based evaluation method should be used for the selection of geometric parameters.
Last Update: 2021-05-31