A Nomogram to Estimate Required Ci-S in Prostate HDR Brachytherapy
J Nasehi Tehrani1*, G Lasio1, B Libby2, N Lamichhane1, K Jiang1, M MacFarlane1, M Guerrero1, (1) University of Maryland, Bel Air, MD, (2) University of Virginia, Charlottesville, VA,
Presentations
PO-GePV-T-1 (Sunday, 7/25/2021) [Eastern Time (GMT-4)]
Purpose: To improve patient safety by providing an independent mathematical model to estimate Curie seconds in prostate HDR brachytherapy.
Methods: HDR prostate brachytherapy, particularly for boost treatment with a single fraction of 15Gy, has become a regular procedure in recent years. For a given source activity, the time needed to deliver the prescribed dose is correlated with the target volume (TV). We calculate the Curie seconds (Ci-s) for each patient plan by multiplying the Iridium-192 source activity in Curies and the total treatment time in seconds. We create a nomogram (Ci-s vs. TV) based on a training dataset of 90 HDR prostate cases with a treatment volume of 15cc to 70cc. Three different curve-fitting models are used in this study (power, polynomial, and logarithmic). We test the models in a validation dataset of 11 additional patients from a different institution and different HDR afterloader vendor.
Results: The coefficient of determination (R2) for the models are 90.45% (power), 90.29% (Polynomial), and 89.68% (logarithmic). The average absolute percentage errors for 11 patients are 4.4% (power), 4.6% (polynomial), and 9.2% (logarithmic). The power curve fit shows the smallest estimation error on 11 patients with a maximum error of 12.4% for 6.84CC and a minimum error of 0.5% for 14.8CC TV. The power curve model for the estimation of Ci-s is 704.39x(TV)^0.5342. The polynomial model parameters are A2=-0.8245, A1=133.66, A0=1115.7, and the logarithmic model is 2433.2xln(TV)–3839.
Conclusion: This method can be used clinically as a sanity check by physicists. The power curve fitting shows the best performance for typical prostate volumes(25-40cc). More extensive clinical tests will finalize the determination of the model for clinical use. Future work will incorporate more variables into the training data to improve the predictive power. We are also going to increase more training data for TV less than 10cc.
ePosters
Keywords
Brachytherapy, HDR, Ir-192 Source
Taxonomy
TH- Brachytherapy: HDR Brachytherapy
Contact Email