The clinical utility of predictive and/or prognostic machine learning models using routinely acquired imaging has resulted in a surge of radiomics and radiogenomics research. Using these methods, large numbers of quantitative imaging features can be extracted in a high-throughput manner, with subsequent feature selection strategies used to systematically find a subset with high predictive power toward a specific task (e.g. survival prediction). While these approaches have traditionally relied upon the use of handcrafted imaging features, automatic feature learning via convolutional neural networks has become increasingly common due to the recent success of deep learning based methods in image-related tasks. In this review, we first present an overview of both the traditional and newer deep learning based radiomics methodologies. Further, we highlight some recent applications of these methods to neuro-oncology.
展开▼
