Transfer Learning-Based Approach for Identification of COVID-19

Corona virus Disease (COVID-2019) spread fast throughout the world, has infected millions of persons, and caused many fatalities. Mobilization has begun throughout the world for this pandemic that is still in existence, with certain constraints and measures being taken to keep this illness from spreading. Furthermore, to manage the illness, affected persons should be found. However, because of the inefficient amount of RT-PCR testing, chest computed tomography (CT) is a common means of supporting COVID-19 diagnosis. The notion of transfer learning was used in this work to detect the covid-19 from the X-ray pictures of the human body chest. With a total accuracy of 92% of the entire model, our model gives the identification of the Covid-19, 96% accuracy. The EfficientNet model previously trained on the Image-Net dataset is used in this study. This research study has customized the changes to the pre-trained model to fit our study and also added a pair of dense and dropout layers before the output layer.

I. Introduction

With the COVID-19 outbreak, the largest pandemic more than 200 countries over 176,945,596 have been infected and over 3,836,828 have died [1]. Although clinicians diagnose the disease clinically, the identification of this sickness, a condition that spreads rapidly, is essential for combating the disease with the early use of data science approaches. According to the investigation by Huang et al. [2] and Singhal et. al. [3], the COVID-19 clinical symptoms include fever, toxins, pharynx, cough, loss of smell and taste. Clinical testing of findings is regarded as the diagnostics reference standard, however, it takes time and leads to misdiagnosis. In addition, the number of COVID-19 test kits in hospitals is restricted. So rapid alternative diagnostic approaches to prevent spread of COVID-19 is essential. At this stage, the known method of early and fast testing medical imagery such as X-ray and computed tomography (CT), employed in chest radiological imaging. Diagnosis is an important function in the treatment of illness of COVID-19 [4]. Studies done by Zhou et al. [5] and Chung [6] et. al. in radiology has focused on the properties of COVID-19 chest CT scans. The initial stage in COVID-19 detection is to come at diagnoses with solid guidance images of radiology for detecting the COVID-19 has been provided by Li et a. [7], Manoharan [24] & Sungheetha [25]. In particular, computer vision is applied for evaluating pictures in deep learning using convolution Neural Networks (CNNs).