Deep Learning Based Gastro Intestinal Disease Analysis Using Wireless Capsule Endoscopy Images

Accurate detection of gastrointestinal illnesses is decisive for early cancer diagnosis and its treatment. However, manual analysis is time-consuming and requires a professional gastroenterologist. An efficient, robust and light-weight multi-class classification framework is proposed for screening different gastrointestinal diseases. A shallow neural network is developed that can extract the discriminative features by convolution of wireless capsule endoscopy (WCE) image even though the diseased images share common patterns. The network is optimised with various optimisation techniques to get the most optimised classification network. The proposed framework is capableof handling the challenges present in the dataset to improve the efficacy of the classification network. The network diagnoses unseen WCE image with 90% accuracy. The developed architecture is compared with other state-of-the-art networks and found to be highly efficient. The proposed network has the potential to perform better in limited computation and resource requirements.