ABSTRACT

The use of quantitative imaging to characterize tumor phenotype, which is correlated to tumor genotype and clinical outcome, is an emerging field of radiology research. Quantitative imaging technologies can facilitate high-throughput mining of quantitative features, a.k.a. radiomic features, from standard-of-care medical imaging modalities such as Computed Tomography (CT), Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET). Radiomics-based decision-support system for precision diagnosis and treatment has shown promise in the field of oncology. In this chapter, we will start with the definitions and implementations of commonly used radiomic features, then focus on the effects of CT imaging acquisition and tumor segmentation on radiomic features and feature-derived predictive models. Finally, we will address the importance of imaging harmonization for robust feature extraction and introduce deep convolutional neural network (CNN), an emerging technique that has the potential to be the ultimate solution for computer-aided cancer diagnosis, prognosis, and response assessment using quantitative medical imaging.