SC2Net: A Novel Segmentation-Based Classification Network for Detection of COVID-19 in Chest X-Ray Images

Abstract

The pandemic of COVID-19 has become a global crisis in public health, which has led to a massive number of deaths and severe economic degradation. To suppress the spread of COVID-19, accurate diagnosis at an early stage is crucial. As the popularly used real-time reverse transcriptase polymerase chain reaction (RT-PCR) swab test can be lengthy and inaccurate, chest screening with radiography imaging is still preferred. However, due to limited image data and the difficulty of the early-stage diagnosis, existing models suffer from ineffective feature extraction and poor network convergence and optimisation. To tackle these issues, a segmentation-based COVID-19 classification network, namely SC2Net, is proposed for effective detection of the COVID-19 from chest x-ray (CXR) images. The SC2Net consists of two subnets: a COVID-19 lung segmentation network (CLSeg), and a spatial attention network (SANet). In order to supress the interference from the background, the CLSeg is first applied to segment the lung region from the CXR. The segmented lung region is then fed to the SANet for classification and diagnosis of the COVID-19. As a shallow yet effective classifier, SANet takes the ResNet-18 as the feature extractor and enhances highlevel feature via the proposed spatial attention module. For performance evaluation, the COVIDGR 1.0 dataset is used, which is a high-quality dataset with various severity levels of the COVID-19. Experimental results have shown that, our SC2Net has an average accuracy of 84.23% and an average F1 score of 81.31% in detection of COVID-19, outperforming several state-of-the-art approaches.

LDDNet: A Deep Learning Framework for the Diagnosis of Infectious Lung Diseases

“…Next, a balanced X-ray dataset was generated from [ 58 ], and the proposed LDDNet was applied to the dataset. This dataset is presented in Table 12 .…”

“…Compared to other activation functions, the computational costs of ReLu are low, and the gradient convergence is also good. For the negative input, ReLu provides zero output and for the positive input, the output is the same as the input [ 52 , 57 , 58 ]. The mathematical equation for ReLu is: …”

Segmentation-Based Classification Deep Learning Model Embedded with Explainable AI for COVID-19 Detection in Chest X-ray Scans

“…Their model performed with an accuracy of 90.3% and an AUC of 0.96. Fang et al [ 55 ] applied a novel CLseg model for segmentation and achieved the Dice of 94.09%. After the segmentation, they applied a novel SC2Net model for the two-class classification of the COVIDGR 1.0 dataset and achieved an accuracy of 84.23% and an AUC of 0.94.…”

“…The following different matrices were utilized for the performance evaluation naming: accuracy, loss, Jaccard index, Dice coefficient, area error, and AUC. The mathematical representations for the matrices are given in the equation below [ 37 , 55 , 56 ]: where TP: True Positive, TN: True Negative, FP: False Positive, FN: False Negative, TP represents the number of COVID-19 samples correctly classified as COVID-19, while FN represents the number of COVID-19 samples wrongly classified to other class. Similarly, TN represents the number of other class samples correctly classified to respective class while FP represents the number of other class samples wrongly classified as COVID-19.…”

A graph‐based edge attention gate medical image segmentation method

“…Due to its excellent performance in semantic segmentation, deep learning has been widely applied in medical image analysis and assisted diagnosis [24]. Its applications include main coronary artery segmentation [25], COVID-19 lung lesion segmentation [26], prostate gland segmentation [27], brain tumor segmentation [28,29], and melanin skin disease segmentation [30]. As a common structure used in medical image segmentation, U-Net combines low-level feature information with high-level feature information through skip connections from encoder to decoder and has achieved better segmentation results with a limited number of data samples.…”