GI bleeding detection in wireless capsule endoscopy images based on pattern recognition and a MapReduce framework

With the goal of relieving the workload of pathologists, recently, the development of computerized approaches for gastrointestinal (GI) bleeding detection in wireless capsule endoscopy (WCE) videos has become an active area of research. Existing methods suffer from either low diagnostic accuracy or high computational cost. In this paper, we present an automated bleeding detection strategy that first discriminate bleeding frames from the normal ones, with further segmentation of the bleeding regions using pattern recognition approaches. The proposed method is then adapted to a MapReduce framework to achieve distributed processing, which dramatically reduces the computing time. The experimental results demonstrate that our method performs a high level of bleeding detection while simultaneously reducing the time needed for computational processing, with achieving up to 0.9804 classification accuracy (in terms of F
₁
score) and 0.8498 segmentation precision.