Deep Learning Framework for Covid-19 Detection and Severity Classification towards Clinical Decision Support System

Abstract

Chest CT scans are widely used for COVID-19 diagnosis. Existing methods focused more on the detection of the disease. However, there is need for detection of severity towards making decisions for suitable course of action. Towards this end, we proposed a deep learning framework for automatic COVID-19 diagnosis and severity detection. Our framework is based on enhanced Convolutional Neural Network (CNN) model which is found efficient for medical image analysis. We proposed two algorithms to realize the framework. The first algorithm is known as Deep Learning based Automatic COVID-19 Diagnosis (DL-ACD). This algorithm is meant for diagnosis of COVID-19 with learning based phenomena. The second algorithm is known as Automatic COVID-19 Severity Detection (ACSD). It is designed to know severity of the disease which helps in making treatment appropriate. Our framework is evaluated against existing deep learning models and found to have superior performance over the existing models.