Click here to


Are you sure ?

Yes, do it No, cancel

A Knowledge-Based Organ Dose Prediction Tool for Brachytherapy Treatment of Cervical Cancer Patients

T I Yusufaly*, A Simon, S Hild, D Brown, D Scanderbeg, J Mayadev, K L Moore, S M Meyers, University of California, San Diego, La Jolla, CA


(Monday, 7/15/2019) 1:15 PM - 1:45 PM

Room: Exhibit Hall | Forum 6

Purpose: The purpose of this work was to explore the accuracy of knowledge-based organ-at-risk (OAR) dose estimation for high-dose rate (HDR) brachytherapy treatment of cervical cancer. Using established external-beam knowledge-based dose-volume histogram (DVH) estimation methods, we sought to accurately predict bladder, rectum, and sigmoid D2cc for standard tandem-and-ovoid (T&O) treatments.

Methods: 32 (16:16 training:validation) locoregionally advanced cervical cancer patients treated with 136 (70:66) T&O CT-guided HDR brachytherapy fractions were analyzed. Single fraction prescription doses were 5.5-7Gy, with external beam+brachytherapy equivalent dose in 2Gy (EQD2) planning goals for high-risk clinical target volume (HRCTV D90>85Gy) and OARs (D2cc_bladder<90Gy; D2cc_rectum<75Gy; D2cc_sigmoid<75Gy). DVH estimation models were obtained by subdividing OARs into HRCTV boundary distance sub-volumes and computing a cohort-averaged differential DVH estimate from the training set sub-volumes. Full DVH estimation was performed on all cases in the training and validation sets by applying OAR sub-volume DVH models to each fraction’s OARs, with model performance quantified by ΔD2cc=D2cc_predicted-D2cc_actual (mean and standard deviation). ΔD2cc’s between training and validation sets were compared with an unpaired Student’s t-test (p>0.05 significance threshold).

Results: Observed D2cc [minimum, median, maximum] in the training:validation sets were D2cc_bladder(Gy)=[1.78, 4.41, 5.94 training; 2.29,4.54,6.37 validation], D2cc_rectum(Gy)=[1.35, 2.96, 5.28 training;1.12, 3.38, 5.69 validation] and D2cc_sigmoid(Gy)=[1.84, 3.90, 5.17 training; 1.77, 3.94, 6.58 validation]. Training set deviations were ΔD2cc_bladder=0.02±0.48Gy, ΔD2cc_rectum=-0.18±0.50Gy, and ΔD2cc_sigmoid=-0.06±0.48Gy. Within the validation set, model accuracy was statistically identical: ΔD2cc_bladder=0.04±0.62Gy (p=0.55), ΔD2cc_rectum=-0.17±0.54Gy (p=0.32), and ΔD2cc_sigmoid=-0.09±0.65Gy (p=0.60).

Conclusion: A simple boundary distance-driven knowledge-based DVH estimation exhibited promising results in predicting critical brachytherapy dose metrics. Incorporating geometric quantification beyond this single-variate formalism will likely increase accuracy.

Funding Support, Disclosures, and Conflict of Interest: KLM acknowledges funding support from AHRQ (R01 HS025440-01). KLM acknowledges research funding, travel support, and honoraria from Varian Medical Systems. SM, JM, and KLM acknowledge support from Padres Pedal the Cause.


Brachytherapy, Treatment Planning, Quality Control


TH- Brachytherapy: Dose optimization and planning

Contact Email