Extended VGG16 Deep-Learning Detects COVID-19 from Chest CT Images

Coronavirus disease 2019 (COVID-19), is a rapidly spreading disease that has infected millions of people worldwide. One of the essential steps to prevent spreading COVID-19 is an effective screening of infected individuals. In addition to clinical tests like Reverse Transcription-Polymerase Chain Reaction (RT-PCR), medical imaging techniques such as Computed Tomography (CT) can be used as a rapid technique to detect and evaluate patients infected by COVID-19. Conventionally, CT-based COVID-19 detection is performed by an expert radiologist. In this paper, we will completely and utterly discuss COVID-19. We present a deep learning Convolutional Neural Network (CNN) model that we have developed to detect chest CT images with COVID-19 lesions. Afterwards, based on the fact that in an infected individual, more than one slice is involved, we determine and apply the best threshold to detect COVID-19 positive patients. We collected 5,225 CT images from 130 COVID-19 positive patients and 4,955 CT images from 130 healthy subjects. We used 3,684 CT images with COVID-19 lesions and their corresponding slices from healthy control subjects to build our model. We used 5-fold-cross-validation to evaluate the model, in which each fold contains 26 patients and 26 healthy subjects. We obtained a sensitivity of 91.5%±6.8%, a specificity of 94.6%±3.4%, an accuracy of 93.0%±3.9%, a precision of 94.5%±3.5%, and an F1-Score of 0.93±0.04.