Hybrid Gradient Descent-tuned Recurrence Net-based Classification of COVID-19 using Chest X-ray images

Abstract

The COVID-19 pandemic has underscored the importance of rapid and accurate diagnosis to mitigate its spread. Chest X-ray (CXR) images have proven to be invaluable tools for identifying pulmonary abnormalities associated with COVID-19. We propose a novel Hybrid Gradient Descent-tuned Recurrence Net (HGD-TRN) approach based on deep learning (DL)regarding the categorization ofCOVID-19. We collected a “dataset of CXR images, including COVID-19 and images of healthy lungs” from Kaggle. The distinction between COVID-19 and healthy patients in this study was made using CXR images. Using the Min-Max Normalization technique, the gathered CXR images are pre-processed. After that, we can obtain images of the lungs by utilizing the Edge-Based Segmentation, which can be used to delineate object boundaries. Next, significant points were extracted from the segmented lung images using Principal Component Analysis (PCA) is an area that uses medical imaging to extract several statistical components, including intensity, shape and texture features. Finally, we usedthe Hybrid Gradient Descent-tuned Recurrence Net (HGD-TRN) method to classify the images into COVID-19 and healthy lung images. The proposed model's performance isevaluated using the Python platform and compared to that of the existing used COVID-19 detection methods. The accuracy 98%, F1 scores 98%, precision 99%,recall 98.90%,Root Mean Squared Error (RMSE) of 0.021and Mean Squared Error(MSE) 0.147, was key examination metrics used to examine the performance of our model. In conclusion, our novel approachHGD-TRN, based on DL techniques, has showcased its potential as an efficient and accurate diagnostic tool for COVID-19 using CXR images.