Detection of common thorax diseases from X-Ray images using a fusion of transfer and statistical learning method

Abstract

An essential component of medical diagnosis is the precise detection and localization of anomalies in X-rays of the chest images. It is urgently necessary to develop the most precise automated model to identify thorax diseases because the number of patients with thorax diseases is rising worldwide. In order to build a reliable prediction model for such tasks, experts will need to manually label a sizable dataset of X-ray images. Nevertheless, more data is needed to build exact models to detect these diseases automatically. As a result, we’re committed to creating a model that detects the anomalies from thorax X-rays automatically, learning from a small amount of X-ray image data that is publicly available and easy to get. To do so, we propose a fusion model by combining transfer learning and statistical learning methods. The comparative reference baseline was significantly outperformed. We show that the detection of thorax diseases can be improved by using our fusion model, allowing quicker diagnosis and treatment.