COVID-19 detection from chest CT images using ensemble deep convolutional neural network

Foysal, M.; Aowlad Hossain, A. B. M.

The whole world is now facing an extensive health crisis for transmittable novel Coronavirus Disease (COVID-19) virus from early 2020. One of the main challenges for clinicians is accurate identification of COVID-19 infected patients and then isolating them to prevent the community spreading. Therefore, detection of COVID cases is of great importance. In this paper, we proposed an ensemble deep convolutional neural network (CNN) approach for the diagnosis of COVID-19 cases from the chest computed tomography (CT) images. Three deep CNN models are designed, trained and validated with CT images by tuning the various hyperparameters of the models. Combining the prediction values of these models to enhance the prediction accuracy reducing the overall false prediction rate, ensemble hard voting rules is applied to classify the COVID and Non-COVID cases. Experimental results achieved the overall accuracy of 96% and sensitivity of 97%. The proposed model can make predictions from the CT images within seconds with higher precision which can be a faster option comparing to the usual reverse transcription polymerase chain reaction (RTPCR) technique. © 2021 IEEE.