S56
Brachytherapy - Gynaecology
ESTRO 2026
Digital Poster 1596
predicted by the model: A/ Acceptable without modifications, B/ acceptable after minor corrections, and C/ not acceptable for clinical use. Results: The mean DSC and HD95 for all contours are summarized in Table 1, together with the results of the qualitative review. Three out of four OAR had a mean DSC ≥ 0.80 in the combined test dataset. Clinical acceptability (A + B scores) among the 2 experts reached 94% for bladder, 100% for rectum, 86% for sigmoid, and 92% for bowel. Figure 1 shows an example contour using the deep-learning algorithm compared to the ground truth contours.
Are current MRI-guided brachytherapy techniques sufficient for large cervical tumors? A dose–volume analysis in a large cohort Grégoire LAURENT DELANGHE, Diane SAYADI, Tahar LADJIMI, Marie-Cécile LE DELEY, Séverine RISBOURG, Eric LARTIGAU, Abel CORDOBA University department of radiation oncology, Oscar Lambret Comprehensive Cancer Center, Lille, France Purpose/Objective: Given the negative impact of large tumor volume on local control, EMBRACE II recommends D90 HR-CTV escalation. This study evaluated dosimetric outcomes across HR-CTV in a large consecutive cohort treated with chemoradiotherapy and MRI-guided adaptative brachytherapy. Material/Methods: This retrospective cohort included patients with histological proven cervical tumors ≥ 3 cm, ECOG 0–2, no distant metastases, treated with external beam radiotherapy (EBRT) followed by brachytherapy between 2016 and 2021. HR-CTV and D90 values were recorded when available. Total EQD2 D90 HR-CTV values, including EBRT contribution, were compared across predefined volume groups (<20 cm ³ , 20–30 cm ³ , 30–70 cm ³ and ≥ 70 cm ³ ). Continuous variables were compared using the Kruskal-Wallis test and proportions using Chi ² . Results: A total of 628 patients were included. FIGO 2018 stages were: I (10.7%), II (24.9%), IIIA–IIIB (3.2%), IIIC1 (24.4%), IIIC2 (29.8%), and IVA (7.0%). Dosimetric data were available for 582 (92.7%) patients. For tumors <70 cm ³ , D90 HR-CTV remained stable across groups, with median values of 88.0 Gy (range 75.7–99.0) for <20 cm ³ , 86.9 Gy (46.3–103.3) for 20–30 cm ³ , and 87.0 Gy (46.6–106.2) for 30–70 cm ³ . In contrast, tumours ≥ 70 cm ³ (n=18) showed a significant dose reduction, with a median D90 HR-CTV of 78.3 Gy (61.9–83.4) and 88.9% receiving ≤ 85 Gy. Median D90 differed significantly between HR-CTV <70 cm ³ (87.2 Gy) and HR-CTV ≥ 70 cm ³ (78.3 Gy) (p<0.001). Organ-at-risk constraints were respected in all groups. Over the study period, the use of interstitial needles increased steadily, reaching 70% in 2021. Conclusion: Recommended D90 HR-CTV thresholds were consistently achieved for HR-CTV <70 cm ³ , while volumes ≥ 70 cm ³ remained undertreated despite increased use of interstitial techniques. Tumour- reduction strategies prior to brachytherapy may allow dose escalation in this setting. Neoadjuvant systemic approaches, including immunotherapy, require prospective evaluation in patients with large tumor volumes.
Conclusion: The robust AI-driven auto-segmentation solution trained using MRI images from a single center delivers highly accurate, clinically acceptable contours for MR- based brachytherapy treatments. By automating this labor-intensive step, this tool may shorten the critical gap between MRI acquisition and treatment delivery, optimizing the most time intensive step of the treatment workflow. Keywords: brachytherapy, gynecology, auto- segmentation References: [1] Ronneberger O, Fischer P, Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. In: Medical Image Computing and Computer-Assisted Intervention (MICCAI). Springer, Cham; 2015. pp. 234–241.
Made with FlippingBook - Share PDF online