[Paper Review] A cascade network for Detecting COVID-19 using chest x-rays

Dailin Lv, Wuteng Qi+3 more|arXiv (Cornell University)|May 1, 2020

COVID-19 diagnosis using AI13 references35 citations

TL;DR

The paper proposes Cascade-SEMEnet, a two-stage cascade of SEME-ResNet50 and SEME-DenseNet169, to detect pneumonia types and finely classify COVID-19 from chest X-rays, with Grad-CAM visualization.

ABSTRACT

The worldwide spread of pneumonia caused by a novel coronavirus poses an unprecedented challenge to the world's medical resources and prevention and control measures. Covid-19 attacks not only the lungs, making it difficult to breathe and life-threatening, but also the heart, kidneys, brain and other vital organs of the body, with possible sequela. At present, the detection of COVID-19 needs to be realized by the reverse transcription-polymerase Chain Reaction (RT-PCR). However, many countries are in the outbreak period of the epidemic, and the medical resources are very limited. They cannot provide sufficient numbers of gene sequence detection, and many patients may not be isolated and treated in time. Given this situation, we researched the analytical and diagnostic capabilities of deep learning on chest radiographs and proposed Cascade-SEMEnet which is cascaded with SEME-ResNet50 and SEME-DenseNet169. The two cascade networks of Cascade - SEMEnet both adopt large input sizes and SE-Structure and use MoEx and histogram equalization to enhance the data. We first used SEME-ResNet50 to screen chest X-ray and diagnosed three classes: normal, bacterial, and viral pneumonia. Then we used SEME-DenseNet169 for fine-grained classification of viral pneumonia and determined if it is caused by COVID-19. To exclude the influence of non-pathological features on the network, we preprocessed the data with U-Net during the training of SEME-DenseNet169. The results showed that our network achieved an accuracy of 85.6\% in determining the type of pneumonia infection and 97.1\% in the fine-grained classification of COVID-19. We used Grad-CAM to visualize the judgment based on the model and help doctors understand the chest radiograph while verifying the effectivene.

Motivation & Objective

• Motivate rapid COVID-19 screening when RT-PCR resources are limited by leveraging chest X-ray imaging.
• Develop a two-stage cascade network to first classify normal, bacterial, and viral pneumonia, then finely classify viral pneumonia to detect COVID-19.
• Improve performance with global average pooling, SE attention modules, large input sizes, CLAHE, MoEx, and U-Net-based lung masking.
• Provide model interpretability through Grad-CAM visualizations to aid clinical decision making.

Proposed method

• Construct Cascade-SEMEnet consisting of SEME-ResNet50 and SEME-DenseNet169.
• Use larger input sizes with global average pooling to preserve lesion details and increase receptive field.
• Incorporate Squeeze-and-Excitation (SE) attention in ResNet50 and DenseNet169 branches.
• Apply CLAHE to enhance contrast and MoEx data augmentation to fuse features from different pathological categories.
• Use U-Net to segment lung regions to remove non-pathological features during training.
• Employ Grad-CAM to visualize classification bases on chest radiographs.

Experimental results

Research questions

• RQ1 Can a two-stage cascade network accurately differentiate normal, bacterial pneumonia, and viral pneumonia from chest X-rays?
• RQ2 Can the fine-grained viral pneumonia stage identify COVID-19 among other viral pneumonias?
• RQ3 Do data augmentation (MoEx, CLAHE) and attention mechanisms improve pneumonia classification performance on chest X-ray data?
• RQ4 Does lung-area masking via U-Net improve diagnostic accuracy by removing non-pathological features?

Key findings

• Base models (VGG19, ResNet50, DenseNet169) show varying accuracy on pneumonia-type classification, with DenseNet169 performing best among them.
• Adding global average pooling (GAP) and SE attention improves accuracy and ROC metrics across models.
• SEME-ResNet50 achieves peak accuracy and ROC performance among tested configurations (85.6% on test set; AUC 0.904); Grad-CAM visualizations localize decision areas to lung regions.
• SEME-DenseNet169 also shows strong performance (80.3% test accuracy with GAP/SE bag; AUC notated in figures).
• Compared to baseline, SEME configurations provide substantial gains in accuracy and F1-scores for viral pneumonia and COVID-19 subdivision; Grad-CAM confirms model attention aligns with lung regions.