Sandy Napel
Stanford University
Since the first x-ray image was produced over 125 years ago, specially-trained human beings (radiologists) have had the responsibility to interpret medical images and to recommend action based on them. But recently we have entered an era when computers are at least as good, if not better, than humans at image understanding, particularly for well-defined tasks. Recognizing this, Gillies, Kinahan, and Hricak described a new paradigm for radiology, i.e., that “Images are More Than Pictures, They Are Data” (Radiology 2016). The technology for quantitatively describing image features was coined “radiomics” by Phillipe Lambin et al. in 2012. Since then, radiomics has become a prominent component of quantitative medical imaging research, having been employed in many studies linking imaging to diagnosis, prognostication, response assessment and, in the case of cancer, molecular characterization of tumors and their environment. (…) As we enter a world where radiomics analysis tools become integrated into medical image analysis workstations, it is important for researchers and physicians alike to understand the basis for them, how they are implemented, how to use them, and how to interpret their output. And even as we move into the future, where in addition to or instead of premeditated engineering of radiomics features, the features themselves will be discovered by artificial intelligence (also called “deep learning”) techniques. (…)
Reference: Extract of the preface of the book 《FOUNDATIONS of RADIOMICS》2020 edited by Dr. Jie Tian, Director of the Key Laboratory of Molecular Imaging, Institute of Automation, Chinese Academy of Sciences.