Machine Learning algorithms approach for Gastrointestinal Polyps classification

Abstract

In this paper we applied machine learning techniques for Gastrointestinal Polyps classification from colonoscopy video clips and compared our results to other methods and with the results of clinicians with different levels of experience. Machine learning technology allows us to classify tissues that can reduce the waiting time for patients' results. We tested four machine learning algorithms (Support Vector Machine, Random Forest, Random Subspace and Extra-Trees) for classification of the polyps in hyperplastic, serrated and adenoma lesions. We used a dataset in which there are 152 instances with the three types of lesions, for 76 polyps. The best results were obtained by Random Forest algorithm with the accuracy of 87%, and the worst results were obtained by Support Vector Machine with the accuracy of between 63% and 73%.