A Novel Multi-Stage Residual Feature Fusion Network for Detection of COVID-19 in Chest X-Ray Images

Abstract

To suppress the spread of COVID-19, accurate diagnosis at an early stage is crucial, chest screening with radiography imaging plays an important role in addition to the real-time reverse transcriptase polymerase chain reaction (RT-PCR) swab test. Due to the limited data, existing models suffer from incapable feature extraction and poor network convergence and optimization. Accordingly, a multi-stage residual network, MSRCovXNet, is proposed for effective detection of COVID-19 from chest x-ray (CXR) images. As a shallow yet effective classifier with the ResNet-18 as the feature extractor, MSRCovXNet is optimized by fusing two proposed feature enhancement modules (FEM), i.e. low-level and high-level feature maps (LLFMs and HLFMs), which contain respectively more local information and rich semantic information, respectively. For effective fusion of these two features, a single-stage FEM (MSFEM) and a multistage FEM (MSFEM) are proposed to enhance the semantic feature representation of the LLFMs and the local feature representation of the HLFMs, respectively. Without ensembling other deep learning models, our MSRCovXNet has a precision of 98.9% and a recall of 94% in detection of COVID-19, which outperforms several state-of-the-art models. When evaluated on the COVIDGR dataset, an average accuracy of 82.2% is achieved, leading other methods by at least 1.2%. Index Terms-COVID-19, chest x-ray imaging, MSRCovXNet, feature enhancement module, ResNet-18 I. INTRODUCTION O N January 30, 2020, the World Health Organization (WHO) formally announced the novel coronavirus pneumonia (COVID-19) as a global health emergency [1] , and from March 31, 2020, this was declared as a pandemic [2]. With millions of infected cases and deaths reported in the world [3], COVID-19 has rapidly spread to hundreds of countries and regions. As reported in [4], [5], it has caused more deaths, than the previous coronavirus strains, for instance, the Middle East Respiratory Syndrome (MERS) and the Severe Acute Respiratory Syndrome (SARS). By the end of 2020, the CVOID-19 pandemic has taken massive losses, with respect to the population health [6] and economic recession [7],

Effective extraction of ventricles and myocardium objects from cardiac magnetic resonance images with a multi-task learning U-Net

“…With more extracted contextual information, it is anticipated further improved boundary and refined segmentation can be achieved, using other related techniques such as change detection [25], abnormality detection [26], brain tumor segmentation and recognition [27], mammogram recognition [28], corneal injury detection [29] and feature fusion network [30]. In addition, the anatomy and physiology of the heart will also be integrated for improving the model accuracy and efficacy [31].…”

MutliFusionNet: Multilayer Multimodal Fusion of Deep Neural Networks for Chest X-Ray Image Classification

“…Enhancement of features improved accuracy by 12.21%. Moreover, Fang et al [9] provide a multi-stage feature fusion network for COVID-19 detection by enhancing low-level feature maps. They utilise the ResNet-18 with a multi-stage feature enhancement module to extract and enhance these features.…”

Unmanned Aerial Vehicle Visual Simultaneous Localization and Mapping: A Survey

“…Sun et al [31] proposed Gumbel-softmax trick to enable energy-constrained concrete autoencoders for dimensionality reduction in HSI. In [32], a multilevel residual feature fusion network for medical applications is proposed. In Ren et al [33], a multitask learning U-net is proposed for effective segmentation of cardiac images.…”