Ensemble Of Texture And Deep Learning Features For Finding Abnormalities In The Gastro-Intestinal Tract

An endoscopy is a strategy in which a specialist utilizes specific instruments to see and work on the inward vessels and organs of the body. This paper expects to predict the abnormalities and diseases in the Gastro-Intestinal Tract, utilizing multimedia data acquired from endoscopy. Deep Analysis of GI tract pictures can foresee diseases and abnormalities, in its early stages and accordingly spare human lives. In this paper, a novel ensemble method is presented, where texture and deep learning features are integrated to improve the prediction of the abnormalities in the GI tract e.g. Peptic ulcer disease. Multimedia content analysis (to extricate data from the visual information) and machine learning (for classification) have been explored. Deep learning has additionally been joined by means of Transfer learning. Medieval Benchmarking Initiative for Multimedia Evaluation provided the dataset, which includes 8000 pictures. The data is gathered from conventional colonoscopy process. Using logistic regression and ensemble of different extracted features, 83% accuracy and a F1 score of 0.821 is achieved on testing sample. The proposed approach is compared with several state-of-the-art methods and results have indicated significant performance gains when compared with other approaches.