Evaluation of a Free-Form Intensity Based Deformable Contouring Method for Prostate Image-Guided Adaptive Radiation Therapy (IGART): Multiple Observer Comparison
R Ellis1*, B Traughber1, D Kaminsky1, C Kan1, Z Shen2, S Pirozzi2, J Piper2, D Nelson2, A Nelson2, (1) University Hospitals Case Medical Center, Cleveland, Ohio, (2) MIM Software Inc, Cleveland, OhioSU-E-J-102 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: To evaluate a deformable contouring method for prostate CBCT as part of an automated image-guided adaptive radiation therapy (IGART) workflow using manual contours from multiple observers for comparison.
Methods: Three CBCTs were randomly selected from 6 patients with prostate cancer. Prostate, bladder, and rectum contours were manually defined on the first CBCT by a single observer and on the second and third CBCTs by 3 observers. A commercial free-form intensity based deformable contouring method was used to deform contours from the first CBCT to the second and third CBCTs. The Dice coefficients and the absolute differences in measured daily delivered dose values were calculated among the deformed and manual contours. Two-sided t-tests were performed to determine whether the automated deformable method has similar variability to manual contours from multiple observers.
Results: The mean(SD) Dice coefficients calculated between the deformed contours and each of the manual contours were 0.72(0.10), 0.75(0.13), 0.64(0.10) for the prostate, bladder, and rectum, respectively, and were 0.76(0.08), 0.87(0.05), 0.68(0.10), respectively, averaged between multiple observer comparisons. These differences were not significant for the rectum (p=0.095) but were significant for the prostate (p=0.025) and bladder (p<0.001). The absolute differences in measured daily delivered dose values between the deformed and manual contours were not significantly different when compared to the variability for multiple observers for both the prostate and rectum (p > 0.05). The difference was significant for the bladder (p <0.05).
Conclusion: A free-form intensity based deformable contouring method for IGART was shown to have similar variability to that of multiple observers both for delivered dose evaluation of the prostate and rectum as well as contour similarity of the rectum while a significant difference existed for bladder dose evaluation. This method has the potential to show deviations from planned treatment as part of an automated prostate IGART workflow.
Funding Support, Disclosures, and Conflict of Interest: J Piper, D Nelson, and A Nelson are owners of MIM Software Inc. Z Shen and S Pirozzi are employees of MIM Software Inc.