@Article{cmc.2022.024312, AUTHOR = {Majdy M. Eltahir, Ibrahim Abunadi, Fahd N. Al-Wesabi, Anwer Mustafa Hilal, Adil Yousif, Abdelwahed Motwakel, Mesfer Al Duhayyim, Manar Ahmed Hamza}, TITLE = {Optimal Hybrid Feature Extraction with Deep Learning for COVID-19 Classifications}, JOURNAL = {Computers, Materials \& Continua}, VOLUME = {71}, YEAR = {2022}, NUMBER = {3}, PAGES = {6257--6273}, URL = {http://www.techscience.com/cmc/v71n3/46548}, ISSN = {1546-2226}, ABSTRACT = {Novel coronavirus 2019 (COVID-19) has affected the people's health, their lifestyle and economical status across the globe. The application of advanced Artificial Intelligence (AI) methods in combination with radiological imaging is useful in accurate detection of the disease. It also assists the physicians to take care of remote villages too. The current research paper proposes a novel automated COVID-19 analysis method with the help of Optimal Hybrid Feature Extraction (OHFE) and Optimal Deep Neural Network (ODNN) called OHFE-ODNN from chest x-ray images. The objective of the presented technique is for performing binary and multi-class classification of COVID-19 analysis from chest X-ray image. The presented OHFE-ODNN method includes a sequence of procedures such as Median Filtering (MF)-based pre-processed, feature extraction and finally, binary (COVID/Non-COVID) and multiclass (Normal, COVID, SARS) classification. Besides, in OHFE-based feature extraction, Gray Level Co-occurrence Matrix (GLCM) and Histogram of Gradients (HOG) are integrated together. The presented OHFE-ODNN model includes Squirrel Search Algorithm (SSA) for fine-tuning the parameters of DNN. The performance of the presented OHFE-ODNN technique is conducted using chest x-rays dataset. The presented OHFE-ODNN method classified the binary classes effectively with a maximum precision of 95.82%, accuracy of 94.01% and F-score of 96.61%. Besides, multiple classes were classified proficiently by OHFE-ODNN model with a precision of 95.63%, accuracy of 95.60% and an F-score of 95.73%.}, DOI = {10.32604/cmc.2022.024312} }