CNN-Based Multiclass Classification for COVID-19 in Chest X-ray Images
Managing the increasing number of patients requiring first screening can be significantly aided by real-time automated detection of COVID-19. It's feasible that Deep CNN models that have been trained on sufficiently large datasets will emerge as the most promising options for achieving the goal. This study aims to automatically detect and classify COVID-19 and viral pneumonia infections in chest X-ray images using a deep CNN model. Our proposed model relies on multiclass labeling to accomplish our aims. Design and Organization: Using the ImageNet pre-trained weights, the proposed model is built on top of the VGG16 framework. Additional custom layers were used to fine-tune the model and produce a better overall performance that is more specific to the goal. In terms of its subjects and methods, this study uses 15,153 samples in total. There are X-rays of the lungs from patients with COVID-19, those with viral pneumonia, and healthy volunteers. The entire dataset was split into an 80:20 split for the train and test sets before the model was trained. Image preprocessing and augmentation were used to enhance crucial parts of the photos before they were sent to the model in batches. We measure the model's efficacy with accuracy, precision, recall, and the F1 score. The analysis that was performed statistically was. The model's output is compared to the results of other recent research that have set the standard in the field. The proposed model has a 98% accuracy in multiclass classification on the test dataset, as measured by 98% precision, 96% recall, and 97% F1 score. Receiver operating characteristic curve area scores of 0.99 were achieved in all three multiclass classification situations. Finally, the proposed categorization model may show to be highly useful in the first diagnosis of COVID-19 and viral pneumonia patients, especially when dealing with heavy workloads and large volumes.
Volume & Issue
Vol. Volume 6 / Iss. Issue 1