Detection of coronavirus Disease (COVID-19) based on Deep Features and Support Vector Machine

The detection of coronavirus (COVID-19) is now a critical task for the medical practitioner. The coronavirus spread so quickly between people and approaches 100,000 people worldwide. In this consequence, it is very much essential to identify the infected people so that prevention of spread can be taken. In this paper, the deep feature plus support vector machine (SVM) based methodology is suggested for detection of coronavirus infected patient using X-ray images. For classification, SVM is used instead of deep learning based classifier, as the later one need a large dataset for training and validation. The deep features from the fully connected layer of CNN model are extracted and fed to SVM for classification purpose. The SVM classifies the corona affected X-ray images from others. The methodology consists of three categories of Xray images, i.e., COVID-19, pneumonia and normal. The method is beneficial for the medical practitioner to classify among the COVID-19 patient, pneumonia patient and healthy people. SVM is evaluated for detection of COVID-19 using the deep features of different 13 number of CNN models. The SVM produced the best results using the deep feature of ResNet50. The classification model, i.e. ResNet50 plus SVM achieved accuracy, sensitivity, FPR and F1 score of 95.33%,95.33%,2.33% and 95.34% respectively for detection of COVID-19 (ignoring SARS, MERS and ARDS). Again, the highest accuracy achieved by ResNet50 plus SVM is 98.66%. The result is based on the Xray images available in the repository of GitHub and Kaggle. As the data set is in hundreds, the classification based on SVM is more robust compared to the transfer learning approach. Also, a comparison analysis of other traditional classification method is carried out. The traditional methods are local binary patterns (LBP) plus SVM, histogram of oriented gradients (HOG) plus SVM and Gray Level Co-occurrence Matrix (GLCM) plus SVM. In traditional image classification method, LBP plus SVM achieved 93.4% of accuracy.

namely severe acute respiratory syndrome (SARS) and the Middle East respiratory syndrome (MERS). Indications of coronavirus defilement go in earnestness from respiratory complexity like pneumonia, kidney issue and development of fluid in the lungs.
On 11 th February 2020, the WHO Director-General gave an acronym "COVID-19" to these infections were found to be caused by a new coronavirus. In the last two decades, two coronavirus epidemics are observed, i.e. SARS-CoV and MERS-CoV. The first one started in china, spread to twenty-four countries and reported 8000 cases & 800 of deaths. The second one started from Saudi Arabia, reported 2500 cases and 800 deaths. The detail of coronavirus is depicted in Table 1. Among the causing pathogens for respiratory diseases, CoV is become the dangerous one because of its serial interval (5 to 7.5) and reproductive rate (2 to 3) (Nishiura et al., 2020). The CoV belongs to single-stranded RNA viruses (+ssRNA) family mostly observed in animals (Perlman et al., 2009;Chan et al., 2013). The analysis carried out till date, the viruses have no species barrier and can cause severe diseases like MERS and SARS. The coronavirus infection can provoke SARS that is severe enough to be called Acute respiratory distress syndrome (ARDS). In general, estimates suggest that 2% of the population are healthy carriers of a CoV and that these viruses are responsible for about 5% to 10% of acute respiratory infections (Chen et al., 2020b). COVID-19 spreads more quickly than SARS and have symptoms like other coronaviruses. Figure 1 shows the distribution of COVID-19 cases and deaths worldwide, as of 13 th April 2020 (European Centre for Disease Prevention and Control, 2020).  Therefore, developing an automated analysis system is necessary to save medical professionals valuable time.
In the current situation of the rapid spread of COVID-19 many kinds of research have been going on based on clinical features (Yang et al., 2020;Wang et al., 2020;Chen et al., 2020a).
Deep Learning is a combination of machine learning methods that mainly focused on the automatic feature extraction and classification from images. The object detection and classification are the two main tasks where deep learning is applied. The advancement of machine learning has a great benefit for clinical decision making and development computeraided systems (Greenspan et al., 2016;Wahab et al., 2017;Xia et al., 207;Burlina et al., 2017).
As the data set available is in hundreds, which is very small for transfer learning approach and it's leading a question mark about the robustness of the classification model. So, instead of using a pre-trained network as a classifier in transfer learning approach to detect the COVID-19, we choose SVM as the classifier. The SVM classifies Xray images of COVID-19 patient, pneumonia patient and healthy people with the use of deep features extracted from fully connected layer of the pre-trained network.
In this paper, a classification model is developed, which classify the Xray images of COVID-19, pneumonia patient and healthy people. The main objective of this research is to screen out The frontal view includes posteroanterior (PA) and anteroposterior (AP) views. Mostly these views are used to examine the lung diseases. In addition to GitHub, we collected 79 Xray images from Kaggle repository (Kaggle, 2020). Again, another two sets of Xray images (pneumonia and healthy) are collected from (Kermany et al., 2018) to train the classification model to distinguish the COVID-19 from pneumonia patient and healthy people. The Xray images of COVID-19, Bacterial Pneumonia, Viral Pneumonia and Normal are shown in Figure   2. The detail about the dataset is in Table 2.
We use this dataset for deep feature extraction using pre-trained networks such as AlexNet, The main contribution of this article is as follows.
• We prepared a dataset of 381 Xray images, i.e. chest Xray frontal view images of 127 COVID-19, 127 pneumonia and 127 healthy people.
• The deep features of most used 13 number of deep CNN models (AlexNet, VGG16, InceptionResNetV2, DenseNet201, XceptionNet, MobileNetV2 and ShuffleNet) are extracted and used by SVM classifier for detection of COVID-19.
• The proposed method is a three-class problem to classify the Xray images of healthy people, pneumonia patient and COVID-19 patient.
• Finally, a comparative analysis of deep feature plus SVM and traditional image classification method (LBP+SVM, HOG+SVM and GLCM+SVM) is carried out.
The remaining paper is organised as follow. Section 2 discussed the methodology. The experimental results are detailed in section 3. At last, section 4 concludes with future scope.

Methodology
Deep feature extraction is based on the extraction of features acquired from a pre-trained CNN (Lopes et al., 2017).  In this way, the feature matrix is then diminished in POOLING layer.

Results and Discussion
In this study, we examined the performance of classification models for detection of COVID-19 based on eleven number of CNN models. The experimental studies were implemented using the MATLAB 2019a deep learning toolbox. All applications were run on a laptop, i.e. Acer Predator Helios 300 Core i5 8th Gen -(8 GB/1 TB HDD/128 GB SSD/Windows 10 Home/4 GB Graphics) and equipped with NVIDIA GeForce GTX 1050Ti. The measurement of performance of each classifier is measured in terms of Accuracy, Sensitivity, Specificity, False positive rate (FPR), F1 Score, MCC and Kappa. In addition, this experimentation used One-Vs-all approach and linear SVM as the SVM classifier parameter. The results reported in Table   4 and Table 5 are based on the average value of 20 independent executions. The training and testing ratio is 80:20 and adapted randomized selection for training and testing in each execution. The accuracy of different classification models with their mean, minimum and maximum achievable values in 20 independent execution are recorded in Table 4. Also, the other performance measures such as sensitivity, FPR and F1 score of different classification models are recorded in Table 5.

Comparison with other machine learning approach
In image processing and machine learning approach, mostly HOG plus SVM, GLCM plus SVM and LBP plus SVM are applied for image classification. The accuracy of those approaches is given in Table 6.

Discussion and comparison of simulation results
The proposed study used pre-trained CNN models to obtain the best performance for detection of COVID-19. We evaluated the performance results of deep feature extraction based on the

Conclusion
The content of the manuscript about the coronavirus is based on the data available in WHO,