Describir: Content-Based Retrieval of Focal Liver Lesions Using Bag-of-Visual-Words Representations of Single- and Multiphase Contrast-Enhanced CT Images.