An ensemble algorithm based on deep learning for tuberculosis classification

Abstract

In the past decades the field of Artificial Intelligence, and specially the Machine Learning (ML) research area, has undergone a great expansion. This has been allowed for the greater availability of data, which has not been foreign in the field of medicine. This data can be used to train supervised Machine Learning algorithms. Taking into account that this data can be in form of images, several ML algorithms, such as Artificial Neural Networks, Support Vector Machines, or Deep Learning Algorithms, are particularly suitable candidates to help in medical diagnosis. This works aims to study the automatic classification of X-Ray images among patients who may have tuberculosis, using an ensemble approach based on ML. In order to achieve this, an ensemble classifier, based on three pre-trained Convolutional Neural Networks, has been designed. A set of 800 samples with chest X-Ray images will be used to carry out an experimental analysis of our proposed ensemble-based classification method.