ESTRO 2025 - Abstract Book

S3361

Physics - Machine learning models and clinical applications

ESTRO 2025

Fig.2: Model performance across treatment timepoints: (a)74 Gy group, (b) 60 Gy group.

Conclusion: Radiomic features are highly predictive early in treatment, emphasising the importance of adjustments before the halfway mark to effectively address rectal bleeding. Continuous analysis of radiomics allows clinicians to predict when a patient is approaching a threshold of potential toxicity and adjust treatment parameters accordingly.

Keywords: machine learning, radiomics, adaptive radiotherapy

References: [1] Burnet NG, Scaife JE, Romanchikova M, Thomas SJ, Bates AM, Wong E, et al. Applying physical science techniques and CERN technology to an unsolved problem in radiation treatment for cancer: the multidisciplinary 'VoxTox' research programme. CERN Ideasq J Exp Innov. 2017;1:3-12. http://doi.org/10.23726/cij.2017.457 [2] Yang Z, Noble DJ, Shelley L, Berger T, Jena R, McLaren DB, et al. Machine-learning with region-level radiomic and dosimetric features for predicting radiotherapy-induced rectal toxicities in prostate cancer patients. Radiother Oncol. 2023;183:109593. http://doi.org/10.1016/j.radonc.2023.109593.