A Machine Learning-Based Classification of Chest Diseases Using Computed Tomography Images

Abstract

Chest diseases, particularly those caused by deadly viruses, substantially threaten human life worldwide. The coronavirus has emerged as one of the most widespread and lethal of these diseases. Medical professionals are lacking the accurate and timely detection of coronavirus. Artificial Intelligence (AI) provides such methods and techniques by using which accurate and timely detection of coronavirus possible. The primary goal of this research is to develop a system using Machine Learning (ML) methods which can detect coronavirus timely and accurately. Computed Tomography (CT) images of Lungs were taken. All images were processed using gray scale conversion, histogram equalization and image normalization pre-processing techniques. Region of Interest (ROI) based segmentation adopted to extracted features from pre-processed CT images dataset. Optimized features were collected using feature reduction method. ML classifiers such as Random Forest (RF), J.48, Naïve Bayes (NB), Bagging, Adaboost, Decision Tree (DT), SMO, Logitboost, Radom Tree (RT) and Bayes Net (BN) deployed to classify CT lungs images and produced 95.62%, 94.88%, 73.39%, 95.85%, 90.84%, 92.18%, 93.55%, 92.34%, 90.07% and 80.27% accuracy respectively. Experiments showed that Bagging attained highest accuracy of 95.85%. The proposed model can detect coronavirus from CT lungs images using ML methods more accurately and timely.