Detection of COVID-19 from CT Scan Images Using Transfer Learning

The world was devastated by the Coronavirus disease since 2019 and the World Health Organization declared it a pandemic in March 2020. Finding patients with coronavirus infection at the earliest is very critical and extremely significant for the prevention of the further spreading of the disease throughout society. Currently, this is the most popular research topic worldwide. Deep learning (DL)-based techniques are proven to be most efficient to diagnose Coronavirus Disease-19 (COVID-19) from either chest X-ray or Computed Axial Tomography (CAT or CT) scan images by researchers. We have proposed the technique to identify COVID-19 from CT anatomy of chest using powerful pre-trained DL Convolutional Neural Network-based models like ResNet50, EfficientNetB0, and VGG16 by using transfer learning. Extended Covid-19 CT chest images and X-ray scans Dataset is used for evaluation provided by Mendeley and freely available online. Initially, a few image preprocessing steps are applied to input CAT images and then features are extracted by above mentioned pre-trained models. Random forest classifier is used for CAT image classification into COVID-19 infected or normal person. Experimental result shows that VGG16 achieved good accuracy of 90.02%, precision of 96.71 %, recall of 97.05%, and F1 score of 96.87%.