CrodenseNet: An efficient parallel cross DenseNet for COVID-19 infection detection. (August 2022)

• Record Type: Journal Article
• Authors: Yang, Jingdong; Zhang, Lei; Tang, Xinjun
• Is Part Of: Biomedical signal processing and control. Volume 77(2022)
• Publication Date: 2022-08
• Subjects: CNN -- Cross dense connections -- DenseNet -- One-sided soft thresholding transformation; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2022.103775
• Languages: English
• ISSNs: 1746-8094

Abstract

Graphical abstract: An effective CNN, CrodenseNet is proposed for COVID-19 detection. CrodenseNet consists of two parallel DenseNet Blocks, each of which contains dilated convolutions with different expansion scales and traditional convolutions. The cross-dense connections and one-sided soft thresholding are applied to the layers for filtering of noise-related features, and increase information interaction of local and global features. Cross-validation experiments on COVID-19x dataset shows that via CrodenseNet the COVID-19 detection attains the precision of 0.967±0.010, recall of 0.967±0.010, F1-score of 0.973±0.005, AP (area under P-R curve) of 0.991±0.002, and AUC (area under ROC curve) of 0.996±0.001. CrodenseNet outperforms a variety of state-of-the-art models in terms of evaluation metrics so it assists clinicians to prompt diagnosis of COVID-19 infection. Highlights: A parallel cross-dense module composed of dilated convolutions and traditional convolutions is proposed to extract features. The one-sided soft thresholding is employed to remove noise-related features and reduce the impacts on another channel. Two Auxiliary classifiers are applied to increase regularization for reduction of overfitting, and main classifier fuses the features extracted from the sub-classifiers to make final decisions. Abstract: Purpose At present, though the application of Convolution Neural Network (CNN) to detect COVID-19 infection significantly enhance the detection performance and efficiency, it often causes low sensitivity and poor generalization performance. Methods In this article, an effective CNN, CrodenseNet is proposed for COVID-19 detection. CrodenseNet consists of two parallel DenseNet Blocks, each of which contains dilated convolutions with different expansion scales and traditional convolutions. We employ cross-dense connections and one-sided soft thresholding to the layers for filtering of noise-related features, and increase information interaction of local and global features. Results Cross-validation experiments on COVID-19x dataset shows that via CrodenseNet the COVID-19 detection attains the precision of 0.967 ± 0.010, recall of 0.967 ± 0.010, F1-score of 0.973 ± 0.005, AP (area under P-R curve) of 0.991 ± 0.002, and AUC (area under ROC curve) of 0.996 ± 0.001. Conclusion CrodenseNet outperforms a variety of state-of-the-art models in terms of evaluation metrics so it assists clinicians to prompt diagnosis of COVID-19 infection.