Automated-Detection-and-Classification-of-Gastrointestinal-Bleeding-in-Wireless-Capsule-Endoscopy

Use of wireless endoscopy in detecting digestive tract haemorrhages automatically is a giant step forward in the field of medical diagnosis. Gastrointestinal bleeding (GIB) is among the most critical conditions which require prompt and accurate detection so as to prevent the grave complications and mortalities that may follow. Specifically, wireless endoscopy through its capsule endoscopy has provided a minimally invasive way to visualise the entire GI tract. This project proposes an automated detection system of Gastrointestinal (GI) bleeding employing advanced machine learning algorithms with wireless endoscopy technology.

The proposed system utilises deep learning, especially convolutional neural networks (CNNs), for analysing endoscopic images and detecting bleeding sites in high precision and recall rates. The training dataset consists of 2618 bleeding and non-bleeding WCE frames collected from multiple internet resources, datasets with a vast variety and types of gastrointestinal (GI) bleeding throughout the GI tract along with medically validated binary masks and bounding boxes.

The training dataset is passed through a deep learning architecture based Convolutional neural network(CNN) for feature extraction and then the feature map is passed to the Rectified Linear Unit (ReLU) activation function to get the resulting classification into two classes - GI bleeding and No GI bleeding. The test dataset is an independently collected WCE data containing bleeding and non-bleeding frames of more than 30 patients suffering from acute, chronic and occult GI bleeding referred at Department of Gastroenterology and HNU, All India Institute of Medical Sciences, New Delhi, India. Implementation of this automatic detection system can be quite useful for the diagnosis process in terms of reducing the workloads of physicians, eliminating the risk of missed out bleeding regions and providing immediate medical assistance.