Radiomics: Basic Concepts and Clinical Applications for Oncologic Patients

Today, medical imaging interpretation is solely based on subjective radiological opinions that suffer from increased variability and low reproducibility. Including Radiomics in the diagnostic process is expected to result in the improvement of diagnostic accuracy, as well as the prediction of treatment response and access to valuable early prognosis information. Several modules of the Radiomics pipeline will be presented and discussed starting from the first phase, where a clinician needs to formulate the clinical question. A radiologist will then need to identify the relevant data sources (types of imaging modalities), and after that an imaging scientist should make sure that the data is preprocessed in a way to enable it to be exploited by the next phases in the pipeline. The lesion segmentation should be done by more than one radiologist, and the radiomics features should be computed by an Imaging Scientist in order to proceed to the model training and validation phases made by a Data Scientist.