Concurrent Convolutional Neural Networks with Decision Fusion to Diagnose COVID-19 using Chest X-ray Imagery

• Published in: 2021 13th International Conference on Electronics, Computers and Artificial Intelligence (ECAI) pp. 1 - 4
• Main Authors: Ghenea, Gabriela-Loredana, Neagoe, Victor-Emil
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2021
• Subjects: chest X-ray imagery; Computer architecture; Computers; concurrent convolutional neural networks (CCNNs); COVID-19; COVID-19 diagnosis; decision fusion; Deep learning; Geometry; Training; Visualization
• DOI: 10.1109/ECAI52376.2021.9515174

This paper presents a new deep learning classifier model based on the ensemble of two concurrent Convolutional Neural Networks (CNNs). The CNN modules have identical architectures according to Visual Geometry Group Network (VGG-Net) pattern, but they are intentionally trained with asymmetric volumes of training samples. The system uses a decision fusion to increase the classification accuracy. We have applied the proposed decision fusion classifier to COVID-19 diagnosis using chest X-ray imagery. For experiments, we have chosen a balanced dataset containing 5674 training chest X-ray images (2837 belonging to subjects with COVID-19, and the other 2837 corresponding to subjects non COVID-19). We have obtained a maximum accuracy of 95.43% on the test set using decision fusion.