A crucial point in radiotherapy is to correctly identify bone marrow (BM) and implement strategies able to selectively spare it since the radiation dose received by this structure is a predictive factor for hematologic toxicity occurrence. In particular, the portion of BM to be spared is the hematopoietically active BM (actBM), which is the part in charge of blood cell generation. Several methods exist to identify actBM, based on morphological and functional imaging, such as MRI, SPECT, and PET. CT imaging is a widespread exam and all patients have to undergo it before starting radiotherapy. However, no approaches have been proposed to identify actBM using CT. In this project, we are working, in collaboration with the Department of Oncology of Università di Torino, on the development of a Decision Support System for the identification of the actBM from CT images, based on radiomics and machine learning techniques. This tool is based on a first step of feature extraction, consisting of the calculation of variables able to quantitatively describe the characteristics of the anatomical structure of interest. Then, the feature subsets and the classifier parameters are simultaneously optimized using a Genetic Algorithm. The obtained classifier is used to identify, on the CT images, the areas belonging to actBM.