Data-Driven Covid-19 Detection Through Medical Imaging
Αναστάσιος Αρσένος, Andjoli Davidhi, Dimitrios Kollias, Panos Prassopoulos, Stefanos Kollias
Abstract
The COVID-19 pandemic has significantly affected public health and global economies. Medical imaging, and particularly chest 3-D CT scans, have been a valuable means for early diagnosis of infection by COVID-19. As a consequence, there has been great need for classification models that can effectively analyze the CT scans. In this paper, we introduce a new COVID-19 database and a deep learning model that can be appropriately trained and classify the CT scans with high accuracy. The database includes a comprehensive collection of COVID-19-related medical images from chest CT-scans. The deep learning model incorporates convolutional recurrent neural networks and a routing mechanism that implement the classification task. Experiments are presented which illustrate that this model performs very well and outperforms other state-of-the-art models in COVID-19 classification.