key: cord-0976383-yzhe18js
authors: Narayan Das, Nripendra; Kumar, Naresh; Kaur, Manjit; Kumar, Vijay; Singh, Dilbag
title: Automated deep transfer learning-based approach for detection of COVID-19 infection in chest X-rays
date: 2020-07-03
journal: Ing Rech Biomed
DOI: 10.1016/j.irbm.2020.07.001
sha: 98a52fc7a7b0dec567b107140f2d72d6c4c76d80
doc_id: 976383
cord_uid: yzhe18js

The most widely used novel coronavirus (COVID-19) detection technique is a real-time polymerase chain reaction (RT-PCR). However, RT-PCR kits are costly and take 6-9 hours to confirm infection in the patient. Due to less sensitivity of RT-PCR, it provides high false-negative results. To resolve this problem, radiological imaging techniques such as chest X-rays and computed tomography (CT) are used to detect and diagnose COVID-19. In this paper, chest X-rays is preferred over CT scan. The reason behind this is that X-rays machines are available in most of the hospitals. X-rays machines are cheaper than the CT scan machine. Besides this, X-rays has low ionizing radiations than CT scan. COVID-19 reveals some radiological signatures that can be easily detected through chest X-rays. For this, radiologists are required to analyze these signatures. However, it is a time-consuming and error-prone task. Hence, there is a need to automate the analysis of chest X-rays. The automatic analysis of chest X-rays can be done through deep learning-based approaches, which may accelerate the analysis time. These approaches can train the weights of networks on large datasets as well as fine-tuning the weights of pre-trained networks on small datasets. However, these approaches applied to chest X-rays are very limited. Hence, the main objective of this paper is to develop an automated deep transfer learning-based approach for detection of COVID-19 infection in chest X-rays by using the extreme version of the Inception (Xception) model. Extensive comparative analyses show that the proposed model performs significantly better as compared to the existing models.

The first case of novel coronavirus (COVID-19) was found in Wuhan, China in December 2019. It was assumed 30 that this virus has been originated from animals that have zoonotic nature. However, the source of this virus has not been identified [1] . The first person has been infected from the Wuhan market in Hubei Province. Thereafter, it has been spread in China and infected persons of any age across the whole World [2] . The disease caused by this novel coronavirus is known as COVID-19. The World Health Organization (WHO) announced COVID-19 a pandemic on 11

February 2020.The total number of infected cases are 7,949,710 with 434,177 causality and 4,087,348 recovered cases on 15 June, 2020 at 11:27 GMT [3] . The second-largest population country, India has 333,008 confirmed cases with 9,520 causality and 169,689 recovered cases.

COVID-19 is a respiratory disease that is instigated by a novel coronavirus. The common symptoms appear in the infected person are fever, cough, sore throat, and difficulty in breathing [4] . Vanishing of taste, tiredness, aches, and nasal blockage can also be observed in some patients [5] . The duration between contamination and the first indication of 40 symptoms may be extended to 14 days [6] . The infection of this virus is transmitted through the droplets of patients such as coughing and sneezing. If the person comes indirectly or indirectly contact with an infected person, then the contacted person gets infected. The vaccines/drugs of this disease are not available until now. Isolation and social distancing are the only solutions to this infection. Therefore, the early detection of infected persons is required to stop the spread of infection. 45 The most widely used COVID-19 detection technique is real-time polymerase chain reaction (RT-PCR). However, RT-PCR kits are costly and take 6-9 hours to confirm infection in the patient [7] . Due to less sensitivity of RT-PCR, it provides high false-negative results. To resolve this problem, radiological imaging techniques such as chest X-rays and computed tomography (CT) are used to detect and diagnose . In this paper, chest X-rays is preferred over CT scan. The reason behind this is that X-rays machines are available in most of the hospitals. X-rays machines 50 are cheaper than the CT scan machine. Besides this, X-rays have low ionizing radiations than CT scan [9] . COVID-19 reveals some radiological signatures that can be easily detected through chest X-rays. For this, radiologists are required to analyze these signatures. However, it is a time-consuming and error-prone task. Hence, there is a need to automate the analysis of chest X-rays.

The automatic analysis of chest X-rays can be done through deep learning-based approaches, which may accelerate 55 the analysis time. These approaches can train the weights of networks on large datasets as well as fine-tuning the weights of pre-trained networks on small datasets [10]. However, these approaches applied to chest X-rays are very limited to [6] . Hence, the motive of this paper is to develop an automated deep learning-based approach for the detection of infection in chest X-rays.

The main contribution of this paper is: 60 1. To overcome the less sensitivity of RT-PCR, chest X-rays images are used in this paper to detect and diagnose of COVID-19.

2. In this paper, chest X-rays is preferred over CT scan. The reason behind this is that X-rays machines are available in most of the hospitals. Even, X-rays machine is cheaper than the CT scan machine. Besides this, X-rays has low ionizing radiations than CT scan.

various performance metrics such as accuracy, f-measure, sensitivity, specificity, and kappa statistics.

The remaining structure of this paper is organized as follows. Section 2 presents related work. The proposed deep learning-based technique is presented in Section 3. Section 4 discusses the experimental results and discussion. The concluding remarks are drawn in Section 5.

Recently, deep learning techniques have been used for the analysis of chest X-rays in a short period. Due to low ionizing radiations and portability of X-rays, it has been preferred over the chest CT scan [11] .

Wang et al. [12] developed a deep convolutional neural network (CNN) for the identification of COVID-19 cases from chest X-rays. Their model was trained over 13,975 chest X-ray images. The classification accuracy obtained 80 from the model was 98. 9%. Hemdan et al. [13] developed a COVIDX-Net for automatic detection of coronavirus infected persons using chest X-ray images. COVIDX-Net was trained on 50 normal and 25 confirmed COVID-19 cases.

The classification accuracy obtained from COVIDX-Net was 91% for COVID cases. Narin et al. [14] presented three different CNN models such as ResNet-50, Inception-ResNetV2, and InceptionV3 for classification of COVID-19 from the chest X-ray images. ResNet50 provided better classification accuracy of 98% than the other models. Sethy and Medhi et al. [27] implemented deep CNN for detecting and diagnosis of coronavirus infection using chest X-ray. They tested their model on 150 confirmed cases obtained from the Kaggle dataset. The accuracy obtained from their 120 model was 93%. Asif and Wenhui [28] proposed an automatic COVID-19 detection system using chest X-ray images.

They used Inception V3 with transfer learning for detecting infection in the patient's chest. Their model was tested on 1341 normal, 1345 viral pneumonia, and 864 COVID-19 images. This model achieved was 96% classification accuracy.

Loey et al. [29] implemented a GAN based deep learning technique for COVID-19 detection in chest X-ray. They investigated three models such as AlexNet, GoogleNet, and ResNet-18. GoogleNet model attained 80.6% and 99.9% 125 for four and two class cases. Asnaoui and Chawki [30] presented a comparative study of seven different deep learning architectures for detecting the symptoms of COVID-19 in chest X-ray images. These models were trained over 6087 images. Inception-ResNetV2 provided the classification accuracy of 92.18%.

This section discusses the proposed deep transfer learning-based COVID-19 disease detection model. Initially, a 130 deep convolutional network is defined. Thereafter, a deep transfer learning model is defined. Finally, the proposed deep transfer learning-based model is defined. Figure 1 illustrates the layered structure of the deep convolution neural network. Initially, to extract the features of images, convolutional layer is applied by using different masks. It produces the low-level features. An activate function is then used as [31, 32, 33] :

Here, R l defines latent demonstration of k th feature of the layer. F represents an activation function, a k defines l th feature of L r of the former layers. In case of first layer, l th ⊗ channel of input images with L r channels. X l and S l define the coefficients and biases values, respectively, of k th feature map. Rectified linear unit (ReLU) is utilized as an activation function F . It can be defined as:

If F < 0, then, the value of a is 0 and returns a otherwise.

The n-dimensional vector can be decomposed into real numbers ∈ [0, 1]. by using the Softmax function. It can be computed as:

Here, yj defines the input tensor's parameter.

In this paper, a cross-entropy is used as a loss function. It can be defined as:

Here, Z j defines label of i th image. P j shows i th parameter of obtained results of softmax function.

To minimize the size of features space, pooling layer is considered. Average and max pooling operations are utilized to obtain average and maximum values, respectively.

Finally, the fully-connected layer is utilized to map the output to flatten and linearly separable space. Softmax is then utilized to test COVID-19 disease in chest x-ray images.

However, a convolutional neural network may suffer from the under-fitting issue, as many potential features may not be extracted. Therefore, in this paper, an extreme version of the Inception (Xception) model is used. Figure 2 shows the architecture of the xception model (for more details please see [34] ). Figure 3 shows the proposed deep transfer learning based COVID-19 disease testing model for chest x-ray images.

It utilizes deep convolutional neural networks and Xception model to build the model. Table 1 shows training analyses of the proposed and competitive models in terms of accuracy, f-measure, sensitivity, specificity, and kappa statistics. It clearly shows that the proposed model achieves significantly better performance as compared to the competitive models. Even the proposed model shows significantly lesser uncertainty values, which shows the proposed model provides consistent training results. Table 2 shows verification analyses of the proposed and the competitive models in terms of accuracy, f-measure, 165 sensitivity, specificity, and kappa statistics on the testing dataset. It demonstrates that the proposed model achieves significantly good performance as compared to the competitive models. Even the proposed model shows significantly

lesser uncertainty values, which shows the proposed model provides consistent training results.

To overcome the less sensitivity issue with RT-PCR, chest X-rays images were used in this paper to detect and 170 diagnosis of COVID-19. Chest X-rays were preferred over CT scans. As X-rays machines are cheaper than the CT scan machines, therefore, we have preferred chest x-ray images. Besides this, X-rays has low ionizing radiations than CT scan. From the extensive review, it has been reviewed that chest X-rays images of COVID-19 infected patients show some unique patterns and bilateral changes. However, manual COVID-19 testing from chest x-ray images is not an easy task. Therefore, in this paper, an automatic analysis of chest X-rays was achieved using deep learning-based approaches.

Deep transfer learning approaches were able to train the weights of networks on large datasets as well as fine-tuning the weights of pre-trained networks on small datasets. Extensive comparative analyses have been performed to evaluate the performance of the proposed model by using various performance metrics such as accuracy, f-measure, sensitivity, specificity, and kappa statistics. Comparative analyses reveal that the proposed model outperforms competitive models.

In the near future, the initial parameters of the proposed model can be tuned by using various approaches like parallel 180 Strength Pareto Evolutionary Algorithm-II [35] , non-dominated sorting genetic algorithm-III [36, 37, 38], memetic differential evolution, [39, 40, 41] , genetic algorithm [42] , particle swarm optimization [43, 44] , etc. Also, some per-processing techniques can be used to improve the visibility of chest x-ray images such as integrated means filter [45, 46] , gain gradient filter [47, 48] , etc.

Novel 2019 coronavirus structure, mechanism of action, antiviral drug promises and rule out against its treatment

In silico study the inhibition of angiotensin converting enzyme 2 receptor of covid-19 by ammoides verticillata components harvested from western algeria

Will covid-19 pandemic diminish by summer-monsoon in india? lesson from the first lockdown

Early reflection on the global impact of covid19, and implications for physiotherapy

Covid-19 and intracerebral haemorrhage: causative or coincidental?, 195 New microbes and new infections

Covid-densenet: A deep learning architecture to detect covid-19 from chest radiology images

Chest ct for typical 2019-ncov pneumonia: relationship to negative rt-pcr testing

Imaging profile of the covid-19 infection: radiologic findings and literature review

Utilizing pretrained deep learning models for automated pulmonary tuberculosis detection using chest radiography

Classification of covid-19 patients from chest ct images using multiobjective differential evolution-based convolutional neural networks

Covid-net: A tailored deep convolutional neural network design for detection of covid-19 cases from chest x-ray images

Covidx-net: A framework of deep learning classifiers to diagnose covid-19 in x-ray images

Detection of coronavirus disease (covid-19) based on deep features

Covid-resnet: A deep learning framework for screening of covid19 from radiographs

Automated methods for detection and classification pneumonia based on x-ray images using deep learning

Classification of covid-19 in chest x-ray images using detrac deep convolutional neural network

Can ai help in screening viral and covid-19 pneumonia?

Covid-2019 detection using x-ray images and artificial intelligence hybrid systems

Estimating uncertainty and interpretability in deep learning for coronavirus (covid-19) detection

Covid-19 detection using artificial intelligence

Covid-xpert: An ai powered population screening of covid-19 cases using chest radiography images

Deepcovidexplainer: Explainable covid-19 predictions based on chest x-ray images

Covid-19: automatic detection from x-ray images utilizing transfer learning with convolutional neural networks

Automated detection of covid-19 cases using deep neural networks with x-ray images

Automatic detection of covid-19 infection from chest x-ray using deep learning, medRxiv

Automatic detection of covid-19 using x-ray images with deep convolutional neural networks 245 and machine learning

Within the lack of chest covid-19 x-ray dataset: A novel detection model based on gan and deep transfer learning

Using x-ray images and deep learning for automated detection of coronavirus disease

Convolutional neural network based detection and judgement of environmental obstacle in vehicle operation

Three-stage network for age estimation

Deep learning approach for microarray cancer data classification

Xception: Deep learning with depthwise separable convolutions

Parallel strength pareto evolutionary algorithm-ii based image encryption

Multi-objective differential evolution based random forest for e-health applications

An efficient image encryption using non-dominated sorting genetic algorithm-iii based 4-d chaotic maps

Color image encryption using non-dominated sorting genetic algorithm with 265 local chaotic search based 5d chaotic map

Parallel strength pareto evolutionary algorithm-ii based image encryption

Color image dehazing using gradient channel prior and guided l0 filter

Color image encryption approach based on memetic differential evolution

Beta chaotic map based image encryption using genetic algorithm

Multi-objective particle swarm optimization-based adaptive neuro-fuzzy inference system for benzene monitoring

Improved particle swarm optimization based adaptive neuro-fuzzy inference system for benzene detection

Image dehazing using window-based integrated means filter

Dehazing of outdoor images using notch based integral guided filter

Single image defogging by gain gradient image filter

Single image dehazing using gradient channel prior

☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:The authors declare no conflict of interest regarding the publication of this paper.prop