CoroNet: A deep neural network for detection and diagnosis of COVID-19 from chest x-ray images (2004.04931v3)

Abstract: Background and Objective The novel Coronavirus also called COVID-19 originated in Wuhan, China in December 2019 and has now spread across the world. It has so far infected around 1.8 million people and claimed approximately 114,698 lives overall. As the number of cases are rapidly increasing, most of the countries are facing shortage of testing kits and resources. The limited quantity of testing kits and increasing number of daily cases encouraged us to come up with a Deep Learning model that can aid radiologists and clinicians in detecting COVID-19 cases using chest X-rays. Methods In this study, we propose CoroNet, a Deep Convolutional Neural Network model to automatically detect COVID-19 infection from chest X-ray images. The proposed model is based on Xception architecture pre-trained on ImageNet dataset and trained end-to-end on a dataset prepared by collecting COVID-19 and other chest pneumonia X-ray images from two different publically available databases. Results and Conclusion CoroNet has been trained and tested on the prepared dataset and the experimental results show that our proposed model achieved an overall accuracy of 89.6%, and more importantly the precision and recall rate for COVID-19 cases are 93% and 98.2% for 4-class cases (COVID vs Pneumonia bacterial vs pneumonia viral vs normal). For 3-class classification (COVID vs Pneumonia vs normal), the proposed model produced a classification accuracy of 95%. The preliminary results of this study look promising which can be further improved as more training data becomes available. Overall, the proposed model substantially advances the current radiology based methodology and during COVID-19 pandemic, it can be very helpful tool for clinical practitioners and radiologists to aid them in diagnosis, quantification and follow-up of COVID-19 cases.

A Deep Neural Network for COVID-19 Detection: CoroNet

The paper "CoroNet: A Deep Neural Network for Detection and Diagnosis of COVID-19 from Chest X-ray Images" by Asif Iqbal Khan, Junaid Latief Shah, and Mohammad Mudasir Bhat, introduces a Deep Convolutional Neural Network (CNN) model designed for the automated detection of COVID-19 infection from chest X-ray images.

Methods

The proposed model, CoroNet, leverages the Xception architecture, which is pre-trained on the ImageNet dataset. Xception, an advanced form of the Inception architecture, utilizes depthwise separable convolutions and residual connections to enhance feature extraction while optimizing computational efficiency. For the paper, the authors compiled a dataset from two public repositories, acquiring a total of 1,251 images, divided into four categories: Normal, Pneumonia Bacterial, Pneumonia Viral, and COVID-19. The images were resized to 224x224 pixels and balanced using random under-sampling.

Results

CoroNet demonstrates remarkable performance across multiple classification tasks. For the four-class classification problem (COVID-19, Pneumonia Bacterial, Pneumonia Viral, Normal), CoroNet achieved an average overall accuracy of 89.6%. Precision, recall, and F1-score for COVID-19 detection were 93.17%, 98.25%, and 95.61%, respectively. The performance metrics for a three-class classification (COVID-19 vs. Pneumonia vs. Normal) were further improved, yielding an overall accuracy of 95%. For binary classification (COVID-19 vs Normal or Pneumonia), the model reached a highly impressive accuracy of 99%, with precision and recall rates for COVID-19 detection standing at 98.3% and 99.3%, respectively.

Discussion

CoroNet significantly advances the state-of-the-art in radiology-based methods for COVID-19 detection. The model's high accuracy and sensitivity make it a promising tool for aiding radiologists in diagnosis and follow-up care, especially during times when there is a shortage of testing kits and other resources. Importantly, the high recall rate for COVID-19 class implies a low false-negative rate, which is crucial in a clinical setting to minimize missed diagnoses.

While CoroNet shows superior performance, further improvements are to be expected as more training data becomes available. This will likely enhance the model's robustness and make it suitable for production deployment. Additionally, comparisons with other models such as COVID-Net and ResNet-50 indicate that CoroNet's performance metrics are competitive and often superior, particularly in multi-class classification tasks.

Implications and Future Research

The implications of this research are both practical and theoretical. Practically, CoroNet can be integrated into clinical workflows to provide preliminary screenings, potentially reducing the burden on PCR testing in strained health systems. Theoretically, the use of sophisticated architectures like Xception and techniques such as transfer learning underline the effectiveness of deep neural networks in medical image analysis and set a precedent for future research in this domain.

Future research may focus on enhancing the dataset with more diverse and higher quality images, as well as implementing CoroNet in real-world clinical environments to further validate its effectiveness. Moreover, expanding the model to handle additional types of pulmonary diseases could generalize its utility.

In conclusion, CoroNet represents a substantial contribution to the automated detection and diagnosis of COVID-19 infections through chest X-ray images. Its high accuracy, coupled with minimal data pre-processing, positions it as a valuable tool in the ongoing battle against the COVID-19 pandemic. Further clinical studies and expanded datasets will be pivotal in translating these promising results into real-world clinical benefits.

For continued research and development, the authors have made their source code and dataset available on GitHub, fostering community involvement and further advancements in this critical area.