MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Dice Coefficient of Auto Segmentation and Its Implication On Inverse Prostate SBRT Treatment Planning

C Yan*, B Guo, R Tendulkar, P Xia, The Cleveland Clinic Foundation, Cleveland, OH

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: segmentation is a bottleneck for adaptive treatment planning. Success of auto-segmentation is often measured by the dice coefficient, directly comparing the manual contours with the corresponding auto-contours. The correlation between the dice coefficient and dosimetric endpoints that are used to measure the plan quality is not fully investigated. The purpose of this study is to investigate the impact of dice coefficient on inverse prostate SBRT treatment planning.


Methods: images from seven SBRT prostate patients were automatically segmented using atlas based segmentation by a commercially available software. The automatically generated rectum was compared with manual contours by calculating the dice coefficient. Two VMAT plans, one using the automatically generated rectum and another using manual contour were generated using same optimization criteria. Dosimetric parameters (D95, D50 of PTV and D100 of manually drawn rectum) were compared between the two plans.


Results: dice coefficients for rectum varies from 0.68 to 0.86 with mean value 0.79. Differences of D95 of PTV varies from 0.05% to 1% while differences of D50 varies from 0.34% to 0.81%. D100 of manually drawn rectum for both plans vary from 0.11% to 17.3%. There is counter intuitive correlation between the dice coefficient and D100 difference for rectum, dice coefficient for the smallest difference of D100 was 0.68 while dice coefficient for the largest difference of D100 was 0.86.


Conclusion: though the dice coefficient of rectum varies between 0.68 and 0.86, its impact on the PTV dose are generally negligible (<1%). Therefore, additional parameters might be needed to determine the usefulness of auto contours in inverse treatment planning.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email