Abstract
Among radiological imaging data, Chest X-rays (CXRs) are of great use in observing COVID-19 manifestations. For mass screening, using CXRs, a computationally efficient AI-driven tool is the must to detect COVID-19-positive cases from non-COVID ones. For this purpose, we proposed a light-weight Convolutional Neural Network (CNN)-tailored shallow architecture that can automatically detect COVID-19-positive cases using CXRs, with no false negatives. The shallow CNN-tailored architecture was designed with fewer parameters as compared to other deep learning models. The shallow CNN-tailored architecture was validated using 321 COVID-19-positive CXRs. In addition to COVID-19-positive cases, another set of non-COVID-19 5856 cases (publicly available, source: Kaggle) was taken into account, consisting of normal, viral, and bacterial pneumonia cases. In our experimental tests, to avoid possible bias, 5-fold cross-validation was followed, and both balanced and imbalanced datasets were used. The proposed model achieved the highest possible accuracy of 99.69%, sensitivity of 1.0, where AUC was 0.9995. Furthermore, the reported false positive rate was only 0.0015 for 5856 COVID-19-negative cases. Our results stated that the proposed CNN could possibly be used for mass screening. Using the exact same set of CXR collection, the current results were better than other deep learning models and major state-of-the-art works.
Introduction
In December 2019, the novel coronavirus disease (COVID-19) was found in Wuhan Province of China https://www.who.int/docs/default-source/coronaviruse/who -china-joint-mission-on-covid-19-final-report.pdf [1].Unlike the common cold and flu, COVID-19 is much more contagious, and for human’s immune system, it is an absolute unknown. To be more specific, severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS) are two well-known coronavirus diseases, which have huge mortality rates of 10% and 37%, respectively [2, 3]. As of now, COVID-19 affected more than 6,057,853 people with more than 371,166 death cases across the world (dated, June 01, 2020) [8]. The COVID-19 outbreak spread rate is exponential and is faster than other respiratory-related diseases. Researchers are now limited to small amount of data to predict possible consequences using AI-driven tools.
Conclusion and Future Works
On the whole, in this paper, a light-weight CNN-tailored shallow architecture was proposed to detect COVID-19- positive cases using CXRs against non-COVID-19 ones. The experiments were performed on dataset collection of COVID19-positive, pneumonia-positive, and healthy CXRs. To validate its robustness, 5-fold cross-validation protocol was used on both balanced and imbalanced dataset scenarios, where 321 COVID-19-positive cases were used. We have compared the proposed shallow CNN-tailored architecture by taking (a) popular DL tools, such as MobileNet, InceptionV3, and ResNet50; and (b) state-of-the-art works for COVID-19 detection using CXRs, into account. The proposed model outperformed all and is computationally efficient as it requires less number of parameters. As the proposed shallow CNNtailored architecture has no false negatives, it could be used to screen COVID-19-positive cases in chest X-rays.
In future, the network architecture will be explored, aiming to minimize false positive cases. We also plan to augment data that helps avoid an issue we currently face, i.e., lack of sufficient data.
