Room: Exhibit Hall | Forum 4
Purpose: To perform organ-at-risk dose accumulation in HDR brachytherapy treatments of the cervix using rigid image registration of pre-treatment CT-simulation scans, evaluate registration performance and compare to conventional dose accumulation by addition.
Methods: We retrospectively analyzed 67 individual implants from 14 patients treated with HDR brachytherapy. Prescription doses were 30Gy/24Gy with 6Gy/fraction, administered using Utrecht applicators. Patients underwent CT-simulation and image-guided planning before each treatment. We performed rigid registrations of each pre-treatment CT to the first fraction scan by matching the tandem and ovoids. The individual and cumulative (plan sum) doses to the 2cm3 and 0.1cm3 with the highest dose were calculated for bladder, rectum, and sigmoid organs-at-risk (OAR). We then compared the cumulative doses to the sum of individual fraction doses. Registration performance was evaluated by calculating Dice similarity and overlap coefficients.
Results: The average differences and ranges between cumulative and sum of individual 2cmÂ³ doses were 0.1Â±0.4Gy (-2.1 to 2.7Gy) for bladder, 0.2Â±0.7Gy (-4.2 to 6.4Gy) for rectum, and 0.3Â±1.3Gy (-9.9 to 7.3Gy) for sigmoid. The 0.1cmÂ³ dose differences varied more with 1.0Â±1.0Gy (-3.2 to 8.4Gy) for bladder, and 1.0Â±1.2Gy (-6.3 to 12.3Gy) for rectum, and 1.0Â±2.1Gy (-16.5 to 10.1Gy) for sigmoid. Dice and overlap coefficients were calculated to be 0.939Â±0.015 and 0.977Â±0.018 for bladder, 0.345Â±0.012 and 0.521Â±0.018 for rectum, and 0.178Â±0.021 and 0.245Â±0.028 for sigmoid. Dice coefficients are smaller than overlap coefficients due to the filling variability in these organs. Furthermore, the higher variability in position of rectum and sigmoid at each fraction relative to the cervix decreases structural overlap.
Conclusion: Doses to the 2cmÂ³ and 0.1cmÂ³ of highest dose in OARs were consistent between rigid registration-based plan summing and addition of each fractionâ€™s values. Furthermore, rigid image registration over-/underestimates the cumulative doses to the OARs. Performing the more labor-intensive workflow presented here is thus not recommended.