Improved Approach for Effective Covid-19 Virus Detection Using Hybrid Deep Learning Approaches

The COVID-19 pandemic has created an urgent need for fast and accurate diagnostic methods. This study introduces a deep learning approach for classifying chest X-ray images to detect COVID-19, using the LIDC-IDRI dataset. To ensure high-quality and uniform input, pre-processing was carried out according to the guidelines of the Radiological Society of North America (RSNA). The processes used here comprised of resizing, normalization, data augmentation, noise reduction, histogram equalization, and cropping. These are all followed for the sake of standardizing the image quality to make training of the model more reliable. Two parallel architectures of CNNs, namely ResNet-34 and HRNet, have been chosen to take advantage of the benefits of different architectures. ResNet-34 reduces the problem of vanishing gradient, while the HRNet keeps high-resolution features, which improves the quality of classification. Training the two networks in parallel enabled us to enhance both accuracy and robustness in distinguishing positives from negatives of COVID-19 cases. It was performed based on the performance metrics in terms of accuracy, precision, recall, and F1-score for each model, and the experiment proved that our method is capable of classifying COVID-19 cases from chest X-rays. Results demonstrate the significant role of advanced pre-processing and parallel CNN architectures in improving diagnostic precision, highlighting AI's potential to support medical imaging and enhance COVID-19 diagnosis accuracy during a pandemic. This study underlines the importance of structured pre-processing and parallel deep learning models as essential components in developing effective diagnostic tools for large-scale health crises.