Classification and Localization of Multi-Type Abnormalities on Chest X-Rays Images

Abstract

Chest X-ray images are among the most common diagnostic tools for detecting and managing bronchopneumonia and lung abnormalities, such as those caused by COVID-19. However, interpreting these images requires significant expertise, and misinterpretations can result in false negatives or positives. Deep learning techniques have recently been highly effective in analyzing medical images, including chest X-rays. In this study, we propose two deep learning approaches to classify and localize different abnormalities, including COVID-19, on chest X-rays, which include multi-classification and object detection models that can identify and localize the presence of disease as other common abnormalities. The proposed models are trained on a large dataset of chest X-ray images from sick people (including COVID-19 patients) and validated on an independent test set. Compared to single object models, this paper presents an ensemble of models by combining multiple object detection models to detect multiple abnormalities in the chest X-ray images. Our results demonstrate that the proposed method achieved promising results in both multi-classification and localization of abnormalities, including COVID-19, compared to the state-of-the-art methodologies. The proposed methods have the potential to assist radiologists in the diagnosis of the abnormalities on chest X-ray images and provide a more accurate and efficient interpretation, thereby improving patient outcomes and reducing the burden on healthcare systems.

Keywords

• Deep learning
• multi-classification
• localization
• ensemble model
• bronchopneumonia/lung abnormalities