Improved Grey Wolf Optimisation-Based Feature Fusion Deep Neural Network for Chest Disease Detection

Abstract

Introduction: Automated diagnosis of COVID-19 is an emergent need in the domain of medical image analysis. COVID-19 is an obstructive pulmonary disease and its common symptoms include cold, fever and cough. COVID-19 is a communicable disease and due to its transmissible characteristics, it rapidly transfers and
affects a large number of populations. Conventional analysis of chest X-ray (CXR) images plays a significant role in the detection of abnormal lung regions, but it is a time-consuming and complicated task to examine CXR images of thousands of COVID-19 patients for a radiologist. Consequently, there is a requirement of a fast,
accurate, and reliable computer-aided diagnosis system (CAD).
Method: The primary goal of this study is the selection of the most prominent features of input CXR images to improve classification accuracy. Metaheuristic algorithms are always the best choice for solving the issue of feature selection. In order to obtain the optimal feature subset from the extracted deep ResNet50 and MobileNetV2 features set, a dimension learning hunting-based Grey Wolf Optimisation (GWO) algorithm has been proposed in this study.
Results: Experimentation work shows that IGWO selects minimum 823 features and using these features the obtained COVID-19 image classification accuracy is 98.78% which is comparatively more than the accuracy obtained in case of PSO(98.47%) and GWO(97.78%).
Conclusion: The obtained results indicate that the Improved version of GWO provides better classification accuracy as com- pared to the other original versions of GWO and Particle Swarm Optimisation (PSO) feature selection algorithms.

How to cite this article:
Bohmrah M K, Kaur H. Improved Grey Wolf
Optimisation-Based Feature Fusion Deep Neural
Network for Chest Disease Detection. J Commun
Dis. 2024;56(4):43-56.

https://doi.org/10.24321/0019.5138.202469