SCL-UMD at the Medico Task-MediaEval 2017: Transfer Learning based Classification of Medical Images

Detecting landmarks in medical images can aid medical diagnosis and is a widely researched problem. The Medico task at MediaEval 2017 addresses the problem of detecting gastrointestinal landmarks, keeping into consideration the amount of training data as well as the speed of the detection system. Since medical data is obtained from real-world patients, access to large amounts of data for training the models can be restricted. We therefore focus on a transfer learning approach, where we can borrow image representations yielded by other image classification/detection systems and then train a supervised learning schemes on the available annotated medical data. We borrow the state of the art deep learning classification schemes (VGGNet and Inception-V3 networks) to obtain representations for the medical images and use them in addition to the provided set of features. A joint model trained on all these features yields a Matthew’s Correlation Coefficient (MCC) of 0.826 with an accuracy and F1-score values of 0.961 and 0.847, respectively.