Pneumonia Detection in Chest X-ray using InceptionV3 and Multi-Class Classification
V. Rajinikanth, Seifedine Kadry, Robertas Damaševičius, C. Pandeeswaran, Mazin Abed Mohammed, G. Glan Devadhas
Abstract
The lung is a imperative internal organs in human physiology. The abnormality in the lung will cause severe respiratory problems. Pneumonia is a severe lung infection, and early screening and treatment are essential to heal the illness. This research aims to implement a pre-trained InceptionV3 scheme to detect pneumonia in chest X-ray pictures. This scheme consists of the following phases; (i) Image collection and resizing, (ii) Deep-features extraction using InceptionV3, (iii) Feature reduction with firefly algorithm, (iv) Multi-class classification, and (v) Validation. A four-class classifier is employed in the proposed scheme to classify the X-ray into normal, mild, moderate, and severe classes using 5-fold cross-validation. The experimental outcome of the K-Nearest Neighbor (KNN) classifier confirms that this scheme offered a classification accuracy of 85.18% on the considered image database.