| 概要(Abstract)： |
Radiomics is an emerging research field that aims at transforming medical images into mineable data for discovery of image features related to clinical outcomes. In cancer patients, such clinical outcomes include the efficacy of chemotherapy, survival prediction, and furthermore, prognostic predictors for specific genetic mutations and molecular pathways. This is performed by building high-dimensional vectors (radiomic signatures) consisting of image and clinical features and investigating its associations with the clinical outcomes. In this talk, concepts of radiomics in cancer research will be presented, with a focus on the differences between radiomics and machine learning, especially deep learning techniques. Next, applications of radiomics in lung, head and neck and brain cancers will be introduced. Finally, a methodology for characterization of the repeatability and reproducibility of histogram and texture features, which are essential for developing clinically-useful radiomic signatures, will be demonstrated.
|