A deep convolutional neural network for bleeding detection in Wireless Capsule Endoscopy images

Wireless Capsule Endoscopy (WCE) is a standard non-invasive modality for small bowel examination. Recently, the development of computer-aided diagnosis (CAD) systems for gastrointestinal (GI) bleeding detection in WCE image videos has become an active research area with the goal of relieving the workload of physicians. Existing methods based primarily on handcrafted features usually give insufficient accuracy for bleeding detection, due to their limited capability of feature representation. In this paper, we present a new automatic bleeding detection strategy based on a deep convolutional neural network and evaluate our method on an expanded dataset of 10,000 WCE images. Experimental results with an increase of around 2 percentage points in the Fi score demonstrate that our method outperforms the state-of-the-art approaches in WCE bleeding detection. The achieved Fi score is of up to 0.9955.