Multiclass X-Ray Image Classification Using Transfer Learning Based Deep Neural Network

In recent past years, the world has been surprised and troubled by the novel coronavirus and its impact on all the fields of life. The coronavirus is a very complex structure difficult to understand and till now it’s growing with new mutations. The virus is very contagious and has affected over 80% of the world's population. Several approaches have been tried for detecting COVID-19, including RT-PCR, however, they are both intrusive and inaccurate. Using radiography in conjunction with a radiologist's expertise is a viable alternative method for detecting COVID-19. To avoid the drawbacks of employing chest X-rays for COVID-19 detection, deep learning methods may be utilized instead. A pre-trained neural network can detect the presence of the virus in chest X-rays by looking for certain viral patterns. X-rays may be analyzed for these patterns to determine if a sample is COVID-positive or -negative, pneumonia-related, or otherwise normal. In this article, we use a neural network that has been pre-trained with 5800 X-ray pictures labeled as covid, pneumonia, or normal in order to identify COVID and pneumonia. The model's performance was evaluated using the following metrics: accuracy (98.50%), precision (98.5), recall (98.34), and F1-Score. The proposed model is compared with another pre-trained model and it outperforms.