Data-driven COVID-19 Prediction Mechanisms: Recent Advancement and Open Issues

Abstract

COVID-19 pandemic has spread throughout the world and has resulted in several casualties since 2019. Moreover, it impacted on every aspect of society including the socio-culture, health, and economic sector of the countries. The accurate and timely detection of covid-19 disease may help to tackle the problem of further spread as well as saving numerous lives. The emergence of a computer vision-based paradigm has facilitated the more effective prediction of covid-19 disease thus making the job of doctors easier. This study explicates a focused investigation of state-of-the-art (SOTA) data-driven models for COVID-19 prediction (COV-19). A novel taxonomy is presented for the classification of COV-19 prediction approaches. This research work analyzes the image datasets with X-ray and CT scans of both the normal and COV-19 patients. A brief illustration of evaluation metrics for automatic COV-19 prediction systems is also expounded. Proposed study indicates that significant success has been achieved for the accurate and efficient detection of the COV-19 disease. However, designing of efficient deep-level models that need comparatively lower training overhead is a major open issue. Besides, insufficient datasets for learning deep convolutional neural networks (DCCN) model is a bottleneck for the researchers. Additionally, the generalization of the data-driven robust model that effectively detects multiple and unknown new variants of COVID-19 is a multifarious open research challenge.

Ähnliche Arbeiten