ESTRO 2025 - Abstract Book

S3774

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

Xu, Y., Ju, L., Tong, J., Zhou, C.-M., & Yang, J.-J. (2020). Machine learning algorithms for predicting the recurrence of stage IV colorectal cancer after tumor resection. Scientific Reports, 10(1), 2519. https://doi.org/10.1038/s41598-020 59115-y

2247

Digital Poster Evaluating secondary cancer risk of prostate radiotherapy treatments using a reparametrized version of Shuryak's model Nicolas M Heumann, Beatriz Sanchez-Nieto, Ignacio Espinoza Physics, Pontifical Catholic University, Santiago, Chile Purpose/Objective: There is an increasing demand for predictive models to estimate secondary cancer risk following radiotherapy, given its potential to improve long-term patient outcome. Among the existing models, the one proposed by Shuryak et al.[1][2] incorporates the most comprehensive radiobiological processes [3] but has not been extensively utilized since its publication. By using updated parameter values and incorporating individualized peripheral dose calculations using P3D and IS2aR [4][5], this study aims to estimate secondary cancer risk of several prostate radiotherapy treatments, based on previously reported average dose to organs. Material/Methods: This model was analyzed, reparametrized, and applied to seven TCP and NTCP equivalent prostate radiotherapy plans previously published by Sánchez-Nieto et al. [6] for a prescribed dose of 77.4 Gy in 43 fractions. The plans included IMRT (inverse and direct planning) and 3D-CRT techniques of Siemens, Elekta, and Varian systems. Excess relative risk (ERR) estimates were calculated for six organs: lungs, stomach, thyroid, colon, bladder, and rectum. Results: Total secondary cancer risks, expressed as ERR, ranged from 2.54 to 2.86 across the plans, reflecting variations in risk between different techniques and equipment. Ranking the plans by ERR revealed statistically significant differences between most consecutively ranked plans (p ≤ 0.0388). However, the comparison between EIMi6 (Elekta IMRT inverse planning, 6 MV) and SIMd6 (Siemens IMRT forward planning, 6 MV) yielded a non-significant result (p = 0.288), suggesting comparable risks for these techniques.