According to the invention, it is provided a novel lesion-agnostic and AI powered radiomics methodology, considering the organ as a whole. By analyzing the organ as a whole from a lesion agnostic perspective, the here proposed methods allow for the extraction of additional imaging information than the ones contained in sheer tumors (and their close vicinity), thus allowing with clinical imaging metadata record (clinical ground truth) for the identification of deep-learned information with higher prediction performance over traditional lesion-centric radiomics. Notably, with either of the ML models as generated according to the present invention, achieved Mean Average Error (MAE) is clearly inferior to ones achieved owing to methods belonging to traditional radiomics.
展开▼