Automatic Detection of Coronavirus Disease (COVID-19) Using X-ray Images and Deep Convolutional Neural Networks (2003.10849v3)

Abstract: The 2019 novel coronavirus disease (COVID-19), with a starting point in China, has spread rapidly among people living in other countries, and is approaching approximately 34,986,502 cases worldwide according to the statistics of European Centre for Disease Prevention and Control. There are a limited number of COVID-19 test kits available in hospitals due to the increasing cases daily. Therefore, it is necessary to implement an automatic detection system as a quick alternative diagnosis option to prevent COVID-19 spreading among people. In this study, five pre-trained convolutional neural network based models (ResNet50, ResNet101, ResNet152, InceptionV3 and Inception-ResNetV2) have been proposed for the detection of coronavirus pneumonia infected patient using chest X-ray radiographs. We have implemented three different binary classifications with four classes (COVID-19, normal (healthy), viral pneumonia and bacterial pneumonia) by using 5-fold cross validation. Considering the performance results obtained, it has seen that the pre-trained ResNet50 model provides the highest classification performance (96.1% accuracy for Dataset-1, 99.5% accuracy for Dataset-2 and 99.7% accuracy for Dataset-3) among other four used models.

Analysis of the Deep Learning Approach for COVID-19 Detection using Chest X-ray Images

The paper "Automatic Detection of Coronavirus Disease (COVID-19) Using X-ray Images and Deep Convolutional Neural Networks", presents a meticulous and technical evaluation of utilizing deep convolutional neural networks (CNNs) for the automated detection of COVID-19 utilizing chest X-ray images. This paper explores the application of transfer learning on five pre-trained models: ResNet50, ResNet101, ResNet152, InceptionV3, and Inception-ResNetV2.

Dataset and Experimental Setup

The dataset consists of 341 COVID-19 patient chest X-rays sourced from an open-access repository on GitHub, accompanied by 2800 normal (healthy) chest X-rays, 2772 bacterial pneumonia, and 1493 viral pneumonia X-ray images from other public databases. The authors structured the dataset into three binary sub-datasets to perform three different classification tasks: COVID-19 vs. Normal (Dataset-1), COVID-19 vs. Viral Pneumonia (Dataset-2), and COVID-19 vs. Bacterial Pneumonia (Dataset-3).

Deep Learning Architecture and Methods

The authors employed deep transfer learning, capitalizing on the knowledge embedded in pre-trained models trained on the ImageNet dataset. The deep learning architectures utilized are well-established models: ResNet50, ResNet101, ResNet152, InceptionV3, and Inception-ResNetV2. Each model functions by converting the chest X-ray images into a processable matrix format and performing feature extraction through convolutional layers followed by pooling and fully connected layers.

For hyperparameters, the learning rate was set to $1 \times 10^{-5}$, the batch size to 3, and the number of epochs to 30. The dataset was subjected to 5-fold cross-validation to ensure robustness and the ADAM optimizer was utilized.

Performance Metrics and Results

The models were evaluated based on five performance metrics: Accuracy, Recall, Specificity, Precision, and F1-Score.

• Dataset-1 (COVID-19 vs. Normal): ResNet50 displayed superior performance with an accuracy of 96.1%.
• Dataset-2 (COVID-19 vs. Viral Pneumonia): The highest classification performance was achieved by ResNet50, presenting an accuracy of 99.5%.
• Dataset-3 (COVID-19 vs. Bacterial Pneumonia): ResNet50 maintained its performance edge, achieving accuracy at 99.7%.

These results illustrate the ResNet50 model's consistent superiority across varied binary datasets, corroborating its robustness and ability to generalize well to different tasks.

Discussion and Implications

The findings underscore the potential of deep learning models, particularly ResNet50, in automating COVID-19 diagnosis from chest X-ray images with high accuracy. The implications for clinical practice are significant, as such automated systems can serve as decision-support tools to radiologists, potentially alleviating the burden on expert human resources and facilitating faster, reliable diagnoses during high-demand periods like a pandemic.

However, notable constraints include the limited dataset size, which can influence the model's generalizability. Future research should thus focus on enlarging the dataset with diverse samples from multiple sources to enhance the models' robustness. Additionally, integrating demographic data and exploring hybrid approaches combining classical image processing techniques with deep learning could further refine diagnostic precision.

Conclusion

Overall, this paper effectively demonstrates the application of pre-trained CNN models in identifying COVID-19 from chest X-rays with high accuracy metrics, especially using the ResNet50 model. This paper sets a precedent for subsequent research aiming to improve and scale AI-based diagnostic tools in pandemic response and healthcare scenarios. Leveraging larger datasets and continual tuning of model architectures can develop even more reliable and widely applicable diagnostic systems.