Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 33: Bioimaging
BP 33.6: Talk
Thursday, March 12, 2026, 16:30–16:45, BAR/0205
Predicting treatment response of tumor spheroids from radiomics analysis of post-treatment dynamics — Pejman Shojaee1,2, Tom Bischopink1, Daria Bolotova1, Sona Michlikova2, Leoni A. Kunz-Schughart2, Steffen Lange1,2, and •Anja Voss-Böhme1 — 1DataMedAssist Group, Faculty of Informatics/Mathematics, HTW Dresden — 2OncoRay - National Center for Radiation Research in Oncology Dresden
Radiomics has significantly advanced radiation oncology by providing quantitative, objective metrics to predict therapeutic efficacy. However, these methods have not been applied to three-dimensional, multicellular tumor spheroids yet, which are the preferred in vitro model for pre-animal, pre-clinical selection of novel, future-oriented treatment modalities. We present an AI-driven predictive modeling workflow to predict long-term tumor spheroid relapse using radiomics data from early post-treatment imaging of spheroids of two human cancer cell lines subjected to radiation therapy and hyperthermia. Our approach integrates multiple feature selection methods and machine learning algorithms for optimal classification performance. A detailed evaluation of the model performance reveals a time gain by early prediction of 2-14 days, while cases of late relapse remain challenging. The presented radiomics-based approach reduces the resource-intensive demands associated with prolonged experimental monitoring and allows accurate prediction for up to three days beyond the observation horizon.
Keywords: radiomics; multicellular tumor spheroids; 3D cancer models; radiation oncology; brightfield microscopy