CovidDiagnosis: Deep Diagnosis of COVID-19 Patients Using Chest X-Rays

As the COVID-19 pandemic threatens to overwhelm healthcare systems across the world, there is a need for reducing the burden on medical staff via automated systems for patient screening. Given the limited availability of testing kits with long turn-around test times and the exponentially increasing number of COVID-19 positive cases, X-rays offer an additional cheap and fast modality for screening COVID-19 positive patients, especially for patients exhibiting respiratory symptoms. In this paper, we propose a solution based on a combination of deep learning and radiomic features for assisting radiologists during the diagnosis of COVID-19. The proposed system of CovidDiagnosis takes a chest X-ray image and passes it through a pipeline comprising of a model for lung isolation, followed by classification of the lung regions into four disease classes, namely Healthy, Pneumonia, Tuberculosis and COVID-19. To assist our classification framework, we employ embeddings of disease symptoms produced by the CheXNet network by creating an ensemble. The proposed approach gives remarkable classification results on publicly available datasets of chest X-rays. Additionally, the system produces visualization maps which highlight the symptoms responsible for producing the classification decisions. This provides trustworthy and interpretable decisions to radiologists for the clinical deployment of CovidDiagnosis. Further, we calibrate our network using temperature scaling to give confidence scores which are representative of true correctness likelihood.