The infectious coronavirus disease (COVID-19) has persisted in having devastating consequences for the lives of human beings all across the world. A fast, affordable COVID-19 screening procedure is needed to identify and isolate affected people, preventing the spread of the disease and ensuring appropriate medical treatment. Recent research reveals that deep learning-based screening of COVID-19 from chest x-ray images may be an alternative to commonly used real-time reverse transcription-polymerase chain reaction (RT-PCR) in circumstances where RT-PCR has time and availability limitations. Therefore, the automatic detection of COVID-19 cases through deep learning is garnering popularity. In this paper, we introduces a novel methodology for automated detection of COVID-19 instances from chest x-ray images, employing a fine-tuned deep convolutional neural network (CNN) approach with transfer learning. We employed three pre-trained deep CNN architectures, specifically Inception V3, DenseNet-121, and MobileNet. These deep CNN architectures were trained using a publicly accessible dataset of COVID-19 chest x-ray images, which was obtained from the Kaggle platform. Data augmentation, such as rotation and zooming, has been used to increase the size of the dataset in order to boost model performance. According to the experimental results, a fine-tuned modified Inception V3, DenseNet-121, and MobileNet model provides an overall accuracy of 98.71%, 98.85%, and 96.70%, respectively. The DenseNet-121 model outperforms state-of-the-art models for COVID-19 diagnosis in terms of overall accuracy, precision, recall, and F1-score metrics. The proposed model can predict from Chest x-ray images with higher precision, making it a faster option than the traditional RT-PCR technique.

Novel coronavirus detection biomedical image classification deep learning pre-trained model feature extraction.