S730
Clinical – Lower GI
ESTRO 2026
treatment clinical and MRI variables. Analyses were effect-size–oriented: Cliff’s δ for continuous features, proportions with risk difference and odds ratio for binary features; ROC analyses reported AUC with 95% CIs; Spearman’s ρ and Cramér’s V supported correlation screening with heat-map visualization.
Conclusion: Within the AIO workflow, the subgroup with better plan quality—i.e., patients eligible for weak supervision—was characterized by a higher proportion of males, more frequent MRF positivity, and slightly greater height. Single-variable discrimination was modest, supporting exploratory use only. Accrual is ongoing, more granular stratification and multivariable modeling are planned. Keywords: Auto-segmentation, Auto-planning, Weak supervision Pathological Response and Outcomes Following Long vs Short Course Radiotherapy in Rectal Adenocarcinoma: A Retrospective Analysis from the Pre-TNT Era Manuela Bermudez Zubiria 1 , Jose David Gonzalez Gomez 1 , Marta Bonet Beltran 1 , Priscila Bernard Contreras 1 , Alejandro Rodriguez Gutierrez 1 , Sara Vazquez Gonzalez 1 , Lucia Tueros Farfan 1 , Luis Ramos Garcia 1 , Montserrat Pampols Felip 2 , Jordi Tarragona Foradada 3 , Julio Idrovo-Gavilanes 4 , Jose Enrique Baquedano Baquedano 1 , Elena Garcia Alonso 1 , Virginia Garcia Reglero 1 , Moises Mira Flores 1 1 Radiation Oncology, Hospital Universitari Arnau de Vilanova, Lleida, Spain. 2 Medical Oncology, Hospital Universitari Arnau de Vilanova, Lleida, Spain. 3 Pathology, Hospital Universitari Arnau de Vilanova, Lleida, Spain. 4 Molecular Oncology, Institut de Recerca Biomedica, Lleida, Spain Digital Poster 3945
Results: Between May and November 2024, 46 patients were accrued; 12/46 (26.1%) met weak-supervision eligibility in AIO workflow. Effect-size–oriented baseline profiling showed small between-group differences for continuous features (Cliff’s δ ), with the largest signal for height ( δ = 0.20; 95% CI − 0.08~0.48). Among binary features, the greatest composition differences were observed for MRF positivity and male sex: MRF (+), risk difference (RD) 0.29 (95% CI − 0.02~0.60), odds ratio (OR) 3.86 (95% CI 0.95~15.71); male vs female, RD 0.27 (95% CI − 0.06~0.59), OR 3.18 (95% CI 0.83~12.28). ROC analysis identified baseline predictors with area under the curve (AUC) ≥ 0.60: tumor location (AUC 0.658; 95% CI 0.513~0.794), sex (0.652; 0.512~0.779), MRF status (0.647; 0.482~0.807), and height (0.637; 0.446~0.809). On correlation screening (Figure 1), both endpoints— weak supervision and plan quality score—showed weak associations with any single baseline feature; the strongest correlates were MRF positivity ( ρ≈ 0.29), male sex ( ρ≈ 0.27), and greater height ( ρ≈ 0.21). Mild positive signals were seen for coronary heart disease, prior pelvic surgery, prior malignancy, and EMVI ( ρ≈ 0.12~0.18), whereas T stage, MRI distance from the anal verge, and overall stage were slightly negative ( ρ≈− 0.04~ − 0.07).
Purpose/Objective: Considering current evidence supporting organ-
Made with FlippingBook - Share PDF online