Deep Neural Networks for Pandemic Hindrance Detection and Screening Replication
Dedeepya Sai Gondi, Vamsi Krishna Reddy Bandaru, Kushwanth Gondi, Deepak Chanda, Lubeck Abraham Huaman Ponce, Mangal Singh
Abstract
The unprecedented outbreaks of global pandemics like the current one of the Coronavirus-19 (known as COVID-19) has stayed the fast, accurate and scalable testing methods needed to detect and prevent transmission of infectious diseases from human to human has come in the limelight. In order to improve the performance of Early Detection and Response, we introduce an Archetype model for Pandemic Hindrance Screening in this paper by applying a Deep Neural Network (DNN) replica. Based on the realization that there are multiple factors contributing to the risk of infection, we train a DNN to accurately predict the risk of getting infected using both synthetic as well as real-world data consisting of medical imaging data, symptom-based data, and demographic data. The feature extraction, normalisation and classification blocks in the proposed framework are optimised with the use of backpropagation and adaptive learning rate. To prove the robustness, the scalability and the flexibility of the model for identifying patterns of infections for affected demographic groups, we simulate a number of epidemic situations. Results obtained as a result of comparison with conventional screening techniques show a significant improvement in the processing time and accuracy of the diagnosis. Moreover, the concept is highly customizable and can easily be combined with IoT based monitoring systems and existing healthcare structures. The implications of the study can be seen in how DNNs can change pandemic response strategies as it would empower the health authorities with a proactive tool to tackle public healthrelated issues more effectively through intelligent datadriven decision making.