Session Item

The extraction of large-scale quantitative data from medical images acquired in clinical routine (i.e. radiomics) has emerged as a gold mine for defining new biomarkers. Several studies have demonstrated its added value and complementarity to clinical and dosimetric variables conventionally used in studies aiming at predicting patient response to treatment or survival. In spite of the enthusiasm of the last ten years, several obstacles such as the non-standardization of image acquisition protocols, coupled with the heterogeneity of the available imagers, the lack of access to large multicenter cohorts, or the difficulty of accessing annotated data (segmentation of the volumes of interest in conventional radiomics), to which are added the sometimes mixed performances, have weakened this research theme, making it difficult to translate into clinical use these laboratory image-based signatures. In this context, this presentation will attempt to address the following questions: are there still blocking points for large-scale radiomics studies? What opportunities for radiomics in 2023? Should we expect clinical trials to personalize patient management in the short term or will radiomics remain the subject of ancillary studies in clinical trials? The state of the art will be illustrated on locally advanced cervical cancer for which biomarkers represent an opportunity for guiding therapeutic trade-offs, particularly at the brachytherapy treatment planning stage.