A Hybrid PSO-GA Optimized Approach for COVID-19 Detection Using CT Scan
Abstract
The rapid global spread of COVID-19 has underscored the urgency of developing efficient diagnostic tools. Although RT-PCR is the gold standard for diagnosis, its time and resource-intensive nature emphasizes the need for alternative approaches. This study introduces a three-step hybrid approach for rapid COVID-19 detection using CT scans to address existing challenges. In the first step, pre-trained convolutional neural networks (CNNs), including VGG-16, ResNet-50, and MobileNet-v2, are utilized to extract critical features from COVID-19-affected lung images. In the second step, a hybrid particle swarm optimization (PSO) and genetic algorithm (GA) optimized approach, called Hybrid PSO-GA, is developed and used to select the optimal features that can increase the accuracy of COVID-19 detection. Finally, the selected features are classified using four distinct models. Remarkable classification accuracies of 99.88% and 99.92% were achieved on the COVIDx-2A CT dataset and the SARS-CoV-2 CT-Scan dataset, respectively. This methodology is novel due to the combination of multiple CNN architectures with sophisticated feature selection algorithms, augmenting the strengths of each method to enhance the accuracy of COVID-19 medical diagnosis, where precision is paramount. The method also exhibits efficiency in the feature selection phase by integrating PSO and GA, utilizing the complementary advantages of both methods.