S66
Brachytherapy - Gynaecology
ESTRO 2026
Python (version 3.9). A stepwise Random Forest regression approach with 10-fold cross-validation was employed using ‘R’ (version 4.4.2) to evaluate model performance and avoid overfitting, progressively refining predictors to identify the most robust configuration. Results:
Metabolism References:
References1. Han K et al. A prospective study of DWI, DCE-MRI and FDG PET imaging for target delineation in brachytherapy for cervical cancer. Radiotherapy and Oncology 2016 -08-12;120(3). 2. Chaudary N et al. Orthotopic xenograft model of cervical cancer for studying microenvironmental effects on metastasis formation and response to drug treatment. CP Pharmacology 2011 -06;53(1). 3. Noonan WT et al. Renal function and glucose transport in male and female mice with diet-induced type II Diabetes Mellitus. Proceedings of the Society for Experimental Biology and Medicine 2000;225(3):221–30. 4. Salazar J et al. Glucosamine for osteoarthritis: Biological effects, clinical efficacy, and safety on glucose metabolism. Arthritis 2014 -02-11;2014:1. Predicting HRCTV D90 Using Pre-HRCTV Shape Features and Interstitial Catheter Number Amani A Chowdhury 1 , Eliana Vasquez Osorio 2 , George Kirby 1 , Gerry Lowe 1 , Mohammed Abdul-Latif 1 , Hannah Tharmalingam 1 , Peter Hoskin 1,2 1 Department of Radiotherapy, Mount Vernon Cancer Centre, Northwood, United Kingdom. 2 Department of Cancer Sciences, University of Manchester, Manchester, United Kingdom Purpose/Objective: The EMBRACE II study demonstrated excellent local control and overall survival across all stages of locally advanced cervical cancer using image-guided adaptive brachytherapy (IGABT) with 74% of patients receive hybrid intracavitary/interstitial (IC/IS) techniques to achieve target doses. However, optimising the minimum dose of radiation delivered to 90% of the high-risk clinical target volume (HCRTV D90) remains complex due to variable HRCTV geometry and Digital Poster 2456 applicator configurations.The purpose of this project was to develop a predictive model and online decision support tool that estimates achievable HRCTV D90 (Gy) using geometric features of the HRCTV and IC/IS catheter number to assist in treatment planning and dose individualisation. Material/Methods: A retrospective database of 143 patients with 569 gynaecological brachytherapy fractions was analysed. HRCTV volume x, y, z coordinates were extracted from the treatment planning system (TPS) using the Eclipse Scripting Application Programming Interface (ESAPI). These coordinates were used to reconstruct the geometric HRCTV, create a binary mask (using the scikit-image package (version 0.22.0)) and extract 14 shape features using PyRadiomics (version 3.1.0) in
Model 4 was trained on 97 patients and 349 HRCTV D90 values, where complete data were available, and demonstrated R ² = 0.82, root mean square deviation (RMSE) = 0.59 Gy, and mean absolute error (MAE) = 0.36 Gy, showing strong agreement between predicted and delivered HRCTV D90. Feature importance identified needle number as the most influential predictor, followed by surface area, sphericity and elongation.An interactive web-based decision support tool was developed, using ‘R’ and shinyapp.io, to allow clinicians to input pre-BT shape metrics and catheter count to estimate HRCTV D90 (Gy) as part of the planning process:https://5xnjq3-amani0aleem- chowdhury.shinyapps.io/HRCTV_D90_Decision_Suppor t_Tool/The tool dynamically integrates EBRT and BT EQD210 contributions to provide estimate total dose delivered to HRCT D90 as per ESGO/ESTRO guidelines.
Conclusion: This pre-treatment HRCTV D90 decision support tool demonstrates modest predictive accuracy based on geometric and implant features. By providing real-time feedback before brachytherapy, it has the potential to assist clinicians in planning IC/IS number and expected HRCTV D90 coverage. However, the model does not currently incorporate applicator configuration, OAR doses, or radiomic factors. Future work will include multi-centre validation and integration of applicator and OAR geometry to further refine this tool. Keywords: Brachytherapy planning, Decision support tool References: Pötter R, et al. The EMBRACE II study: The outcome
Made with FlippingBook - Share PDF online