Project Details
Projekt
Structured prospective validation of published radiomics models and assessment of their applicability for clinical translation
Applicant
Professor Dr. Steffen Löck
Subject Area
Nuclear Medicine, Radiotherapy, Radiobiology
Medical Physics, Biomedical Technology
Medical Physics, Biomedical Technology
Term
since 2023
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 515284204
Numerous radiomics studies have been published for all relevant tumour entities with the aim of enabling personalized oncology. However, few are currently evaluated in clinical trials or even applied in clinical practice. Reasons for this translational gap include limited feature reproducibility from different software implementations, limited feature robustness due to changes in imaging devices and image acquisition, non-standardized image pre-processing and under-reported model development, missing external, multicentre, and prospective validation, limited risk-assessment of the obtained predictions, limited clinical interpretability and accessibility for patients and clinicians. Therefore, in this project, we propose to systematically assess the reproducibility and interpretability of published radiomics models, with the aim of finding candidates for inclusion in clinical trials. In particular, we will (i) externally validate published radiomics models on large available prospective and retrospective datasets of patients with head and neck cancer, brain tumours, rectal cancer and prostate cancer; (ii) assess the robustness and interpretability of validated radiomics models using state-of-the-art methods from artificial intelligence; and (iii) summarize the results in guidelines for future reproducible and interpretable radiomics analyses. Thereby, this project provides an essential contribution to the future development of the research field of radiomics.
DFG Programme
Priority Programmes
Co-Investigators
Professorin Dr. Esther Troost; Alex Zwanenburg-Bezemer, Ph.D.
