Room: AAPM ePoster Library
Purpose: evaluate clinical functions of two kinds of software tools based on the auto-contouring results of organs at risk(OARs) in endometrial and cervical cancer patients.
Methods: better-quality images of endometrial and cervical cancer patients (60 cases each) mapped by senior clinicians including the bladder, the rectum from the anal canal to the sigmoid curve, the bilateral femoral heads, kidneys, leg bones, the spinal cord, the pelvis and the small intestine surrounding the PTV are used to build models for Raystation®©. OARs of another 20 patients are auto-contoured by AiContour®© and Raystation®© software and manual mapped respectively. Overlap index (OI) and Dice similarity index (DSC) and Volume difference (DV) are introduced to evaluate the accuracy of auto-contours. The differences of two softwares are compared using independent-sample t-test.
Results: contours of OARs in endometrial and cervical cancer gotten from AiContour®© are much closer to the manual than Raystation®©'s in most of the indicators of the study. Each OI and DSC indexes calculated by AiContour®© is > 0.8, and the Dv indexes of every organs are < 0.1. The results of Raystation®© are inconsistent, the mean Dv of right-kidney gets the best score, 0.003, the worst mean Dv is bladder, over 0.84. It is demonstrated significant differences in OI and DSC gotten by AiContour®© and Raystation®© software. Except the counter of rectum, there are no significant differences in Dv between two kinds of software.
Conclusion: to Raystation®©, AiContour®© is strongly recommended to use in clinical for auto-contouring OARs in endometrial and cervical cancer. It only needs a little modification before planning and designing. Raystation®© may show advantages in other parts which remained further study.