Deep Greedy Network: A Tool for Medical Diagnosis on Exiguous Dataset of COVID-19

Abstract

The extensive outbreak of COVID-19 has created a worldwide health crisis. Transmission of this disease occurs among people through droplets which causes severe respiratory distress and in turn can also lead to fatal death. At the pinnacle of this pandemic, scientists endeavor to discover the medication for the COVID-19 victims. Artificial Intelligence algorithms, especially, deep learning, on the other hand, is used for the diagnosis of the COVID-19 patients but this requires an enormous radiographic data set to effectively provide an optimized outcome for a particular scenario. This work presents a new technique called ‘Deep Greedy Network’ which will work efficiently with a finite number of datasets. In spite of peculiarity caused due to limited dataset, the anomaly of overfitting and underfitting could be effectively overcome using the proposed algorithm. This, in turn, is simultaneously going to be both cost-effective and efficient. The proposed architecture ensures the efficacious result after the proper judgement of the trained model on the given X-ray datasets of COVID-19 cases.