VISUALIZATION OF CHEST X-RAY IMAGES WITH LIME EXPLANATION FOR COVID-19 AND PNEUMONIA CLASSIFICATION

Abstract COVID-19 disease has taken several millions of lives since December 2019. Pneumonia is a lung infection disease killed many in previous years. Research has been going on for the early detection of COVID-19 and Pneumonia from both CXR (Chest X-ray) and CT (Computed Tomography) images. This paper presents a Deep Neural Network implementation using LIME for three class classifications COVID-19, Pneumonia and Normal. The dataset is taken from Mendeley which contains 5228 chest X-ray images that are categorized into three groups such as COVID-19 (1626 images), PNEUMONIA (1800 images) and Normal (1802 images). Manual examination of Chest X-ray images is time-consuming and complex. LIME technology is useful for interpreting a black box Machine Learning model. It explains every individual prediction. LIME contributes to quality interpretation in the research area. Many researchers have developed models using deep learning techniques but they lack the explanation about the interpretation of the result and they got less accuracy compared to our proposed model. Our proposed model got 98.76% accuracy for the standard dataset of 5228 CXR images which outperforms the state-of-the-art models.