Deep Learning Architectures for Improving Effectiveness of Covid Detection – A Pilot Study

Covid 19 has disrupted people’s lives all over the world. It has also resulted in serious infection, if not predicted at an initial stage. Accurate diagnosis is critical for early diagnosis and to provide appropriate treatment at an earlier stage. Therefore, this work aims at developing a deep learning-based diagnosis system to identify onset of covid at an early stage. In the proposed system a review of existing work in literature and need for covid diagnosis using Artificial Intelligence (AI) techniques is analyzed and presented. Further, three deep learning architectures viz., Convolutional Neural Network (CNN), Deep features from pretrained network classified using Support Vector Machine (SVM) classifier and Transfer learning using Resnet 101 architecture are experimented as part of this pilot study. A publicly available Covid CT image dataset is used to evaluate the proposed system. A considerable improvement in relevant performance metrics is observed in Transfer learning using RESNET 101 architecture. Henceforth the proposed system could be used as an aid to clinicians to accelerate covid 19 diagnosis and provide the necessary treatment.