Classification of Pneumonia Lungs Infected X-Ray Images using Statistical based Features

Abstract

Community-acquired pneumonia is classified as mild, moderate or severe by the doctors, taking into account all the risk of complications. A model is proposed for the pneumonia detection divided into two stages from chest X-Ray picture detecting & classifying the occurrence of pneumonia in different stages from images obtained from various sources such as Rajasthan hospital, Kaggle and other diagnostic centres, Jaipur. Various features are used to obtain informative and relevant data from lungs infected X-Ray images for classification of pneumonia severity in this work. It is tried to achieve a remarkable classification performance by extracting features from various infected chest X-Ray images. Classification takes place in various stages - preprocessing, feature-extraction and classification stage. Various filters are used for pre processing and statistical feature extraction algorithm for image classification. After pre-processing, feature extraction is the next step extracting the features for mild and severe stage of pneumonia applying some statistical based methods. Also the approach proposed is tested with different classifiers for classification stage in the research. The best results are produced when all the features are combined. The system is assessed using a dataset that was created and contains around 226 lungs infected X-ray images in which 102 mild lungs infected X Ray images and 124 severe lungs infected images.