Improved Deep Learning with Metaheuristics Driven COVID-19 Diagnosis on Chest X-Ray Images

Recently, Artificial Intelligence (AI) has undergone significant changes in part of medical image processing. Earlier diagnosis using chest X-ray (CXR) images proved that a key solution in fighting COVID19. Several computer-aided diagnostic (CAD) techniques have been used for assisting radiologists and provide a secondary recommendation for the same. CXR has become the first screening method to play a considerable role in the analysis of COVID19 contamination. Formerly, various traditional machine learning (ML) and deep learning (DL) systems were utilized for automatic diagnosis of chest radiography X-ray images. One method widely applied with DL is transfer learning (TL) which re-uses the data learned from trained modules like resolving one problem and applying it to the same problem. This study presents an Improved Deep Learning with Metaheuristics based disease recognition and classification (IDLM-D2C) for COVID19 on CXR images. The presented IDLM-D2C approach concentrates on the precise detection and classification of COVID19 on CXR images. Initially, the IDLM-D2C algorithm implements the Weiner filter (WF) as a noise elimination step. In addition, the IDLM-D2C method exploits seagull optimization algorithm (SGO) with an improved Faster SqueezeNet model for feature extraction. For COVID19 detection, deep learning modified neural network (DLMNN) model was used with chimp optimization algorithm as hyperparameter optimizer. The experimental validation of the IDLM-D2C method is tested on benchmark medical image dataset. The comparison analysis highlighted the improvements of the IDLM-D2C method over other approaches.