XNET: CNN MODEL for COVID-19 Classification using Skip Connection

Abstract

The dangerous virus so far has been the new Coronavirus A shocking turn of events has resulted in a catastrophic disease that has swept the globe. This lethal outbreak has sparked worries about the burden it puts on healthcare systems, especially when it comes to serving a sizable population that needs urgent care. RT-PCR RAT has emerged as the two main tests used to meet the urgent demand for diagnostics. These tests are preferred because they can produce results quickly. However, relying only on these two approaches has several drawbacks. The issue is made worse by the fact that many countries struggle to obtain enough testing kits. Further complicating the problem are reports of false-positive results. Chest X-rays are a supplemental tool that can be extremely helpful in reducing these difficulties and stopping the spread of COVID. Healthcare practitioners can get quick findings from RT-PCR testing and chest X-rays, allowing patients to take the required precautions immediately. This multifaceted strategy improves the effectiveness of diagnosis and therapy while easing pressure on testing facilities. Utilizing the benefits of various diagnostic techniques, we can address the current issue more skillfully and protect public health. In order to categorize patients with COVID-19 using chest X-ray (CXR) pictures, A CNN Model is first implemented within this research. The Kaggle-sourced dataset for this study consists of two kinds of images: infected and healthy. Techniques for transfer learning are used to improve the CNN. Model’s functionality and effectiveness. We want to enhance the accuracy and reliability of the classification process specifically for COVID-19 identification in CXR pictures by utilizing existing knowledge and skills from pre-trained models. After applying several transfer learning techniques we face the problem of overfitting. To overcome that overfitting, we used residual blocks (skip-connection). After applying this model, we got an accuracy of 96%. So, this technique will help in the future to avoid all problems of expenses and difficulties that we face in hospitals. Overall, this technique has the potential to enhance the accuracy, speed, and efficiency of Covid-19 classification, making it an important tool in the fight against the pandemic.