ABSTRACT
Radiomics is the high-throughput extraction of many quantitative image features with the goal of creating mineable data. Thus, it is a fundamentally big data approach to quantitative imaging. The radiomics workflow can be divided into four stages: the workflow starts with the collection of medical images; the regions of interest are segmented; the radiomics features are extracted; and the radiomics features are used to build models for prediction and inference. The chapter describes each of these stages in more detail, presents some compelling results from radiomics, and discusses the future of radiomics. The first step in the radiomics workflow begins with identifying the characteristics of the patient group to be studied. Depending on the radiomics task, the need for homogeneity in the patient cohort should be carefully assessed. Derivative features are standard features that have been modified to make them more effective for modeling a specific phenotype.
