CNN Based COVID-19 Detection: Enhancing Accuracy, Efficiency and Generalization in Medical Imaging Analysis
Abstract
Medical image analysis through convolutional neural networks (CNN) has emerged as a critical area of research, especially in the field of detecting diseases such as COVID-19. This study focuses on the detection of COVID-19 cases among the normal and pneumonia cases, enhancing the accuracy and efficiently using the capabilities of CNN-based models. The main challenges of this study are data scarcity, class imbalance and the necessity for robust performance across diverse patient demographics and imaging conditions. By utilizing a hybrid model namely, ResNet50 with GRU, this study addresses the key challenges such as image variations and limited available labeled data. The hybrid ResNet50 with GRU model achieves an accuracy of 97.1%, recall of 97.1%, precision of 97.1% and F1 score of 97.1%. Through comprehensive experimentation, our proposed model outperforms other traditional machine learning models in diagnostic accuracy, significantly reducing false negatives and false positives. Moreover, by applying efficient computational strategies this model can function swiftly, enabling it to be suitable for real-time clinical application which can improve patient outcomes and manage public health responses.