Detection and Classification of Lung Diseases Infected by Covid-19 Using FLICM Segmentation and HOG Feature Based SCA-ELM Model

Abstract

Medical imaging achieves enormous claims on automated, reliable, and efficient diagnosis in all medical-related issues. During the present pandemic situation of covid-19, enormous numbers of people are affected by lung diseases. Lung infections are evolved as different types of diseases. It is a problematic task for a clinical physician to segment, detect, and extract infected areas of the lung due to covid-19 and classify the type of infection and its stage from X-ray images. During the present pandemic situation of covid-19, enormous peoples are affected with lungs diseases. Lung infections are evolved as different types of diseases. It is a tough and multifarious the medical practitioner to segment, detect, and extract infected areas of lung due to covid-19 and classify the type of infection and its stage from X-ray images. The size, shape, and position of infection differ from dissimilar patients lung. Several efforts have been proposed for image detection and classification in the literature by considering machine learning and deep learning models. Motivated by the advancement of machine learning for classification of lung diseases, we are proposing a novel segmentation technique and a classifier for detection and classification of lung diseases. This research suggests a HOG (Histogram oriented Graph) feature based hybrid modified SCA (Sine Cosine Algorithm)-Extreme Learning Machine model to classify different infected and non infected lung diseases related to COVID-19. An image enhancement technique also suggested by utilizing Modified Sine Cosine (MSCA) optimization to increase the quality of images. Further, a new Fast and robust Fuzzy Local Information C Means (FRFLICM) segmentation technique for detection of lung infection from X-ray images that can inform radiologist and doctors about the details of lung infection due to covid-19. The lungs chest X-ray images are collected from different hospitals of Ethiopia and Kaggel website. The quality measures SSIM and PSNR are considered for image segmentation and achieved 0.9878 and 35.26 and segmentation accuracy as 98.93%. The performance measure for classification are considered as sensitivity, specificity, classification accuracy and achieved as 98.78%, 99.23%, 99.24% and computational time and archives as 24.1128 seconds The performance comparison the results from the support vector machine and conventional ELM machine learning model with the suggested SCA-ELM models are presented to validate the uniqueness of the research.