Traditional Feature Engineering and Deep Learning Approaches at Medical Classification Task of ImageCLEF 2016 FHDO Biomedical Computer Science Group (BCSG)

Abstract. This paper describes the modeling approaches used for the Subfigure Classification subtask at ImageCLEF 2016 by the FHDO Biomedical Computer Science Group (BCSG). Besides traditional feature engineering, modern Deep Convolutional Neural Networks (DCNN) were used, trained from scratch and using a transfer learning scenario. In addition Bag-of-Visual-Words (BoVW) were computed in Opponent color space, since some classes in this subtask can be distinguished by color. To remove unimportant visual words the Information Gain is used for Feature Selection. Overall BCSG achieved top performance for all three types of features: textual, visual and mixed