ESTRO 2025 - Abstract Book

S2399

Interdisciplinary – Other

ESTRO 2025

patients with breast and head and neck tumors showed a significantly lower risk compared to the global risk of drop-out: OR=0.25, p<0,001 and OR=0.75, p=0.047 respectively. Regarding predictive models, while a custom neural network reached a balanced accuracy of 0.596, its performance remains insufficient for reliably predicting drop outs.

Conclusion This study highlights the importance of structured data in radiotherapy preparation to describe workflow efficiency and patient care. Identifying bottlenecks can streamline processes and reduce delays. However, the predictive model for drop-outs did not yield successful results. Future studies should focus on expanding the dataset, including clinical data, in the hope of improving prediction of the pathway.

Keywords: RO Workflow, Markov Chains, Predictive Models

3603

Digital Poster Real-world scenario from the Department of Radiation Oncology, North India for carcinoma esophagus mrinalini verma, DIVYA KUKREJA, kirti srivastava, rajeev gupta radiation oncology, KGMU, LUCKNOW, India Purpose/Objective A retrospective analysis of histopathologically confirmed oesophageal cancer patients was conducted at a government-funded tertiary cancer care center in North India to analyze the real-world data.