Deep Learning Model for COVID-19 Diagnosis: Improving Accuracy and Sensitivity in Early Detection

The continuous COVID-19 pandemic, caused by the SARS-CoV-2 virus, required fast and efficient diagnostic tools. This work presents a deep learning-based system, using convolutional neural networks, for the detection and diagnosis of COVID-19 through computed tomography tests, aiming to assist specialized medical professionals. A total of 746 Computed Tomography images (CT), were used in this work, one of the largest publicly available chest computed tomography dataset for research into COVID-19. Our proposed technique showed the accuracy of more than 99% for the training set, with high sensitivity and specificity, and achieved 97% on the validation set. Such results would hint at the very possible implementation of our deep CNN approach in clinical diagnostic settings, particularly for COVID-19 testing, to enhance early detection and management for patients.