A Deep Learning Approach for Red Lesions Detection in Video Capsule Endoscopies Paulo Coelho 1(B ) , Ana Pereira 2 , Argentina Leite 3 , Marta Salgado 4 , and Ant´ onio Cunha 3 1 Polytechnic Institute of Leiria, Leiria, Portugal 2 University of Tr´as-os-Montes e Alto Douro, 5000-801 Vila Real, Portugal paulo.coelho@ipleiria.pt 3 INESC TEC (Formerly INESC Porto) and UTAD – University of Tr´as-os-Montes e Alto Douro, 5000-801 Vila Real, Portugal 4 Centro Hospitalar Porto, Porto, Portugal Abstract. The wireless capsule endoscopy has revolutionized early diagnosis of small bowel diseases. However, a single examination has up to 10 h of video and requires between 30–120 min to read. Compu- tational methods are needed to increase both efficiency and accuracy of the diagnosis. In this paper, an evaluation of deep learning U-Net architecture is presented, to detect and segment red lesions in the small bowel. Its results were compared with those obtained from the literature review. To make the evaluation closer to those used in clinical environ- ments, the U-Net was also evaluated in an annotated sequence by using the Suspected Blood Indicator tool (SBI). Results found that detection and segmentation using U-Net outperformed both the algorithms used in the literature review and the SBI tool. Keywords: Lesion detection · Gastrointestinal bleeding Machine learning · Capsule endoscopy · Deep learning · U-Net 1 Introduction Approximately 300,000 hospitalizations per year in the United States of America are associated with gastrointestinal bleeding and in 5% of those cases it is not possible to immediately identify the bleeding’s source [14]. The small bowel is one of the major organs where bleeding from unknown sources occurs (also named as Obscure Gastrointestinal Bleeding - OGIB). Full and direct visualization of the small bowel is not possible through high endoscopy or colonoscopy, due to the organ’s length and its morphological diversity [8]. To overcome this issue, direct visualization of the small bowel through endoscopic methods with emphasis to the Wireless Capsule Endoscopy (WCE), have greatly evolved in the last decades, revolutionizing the knowledge and clinical approach of several pathologies [10]. c  Springer International Publishing AG, part of Springer Nature 2018 A. Campilho et al. (Eds.): ICIAR 2018, LNCS 10882, pp. 553–561, 2018. https://doi.org/10.1007/978-3-319-93000-8_63