Advancing COVID-19 Prediction with Deep Learning Models: A Review
Adetokunbo MacGregor John-Otumu, Charles Ikerionwu, Oluwaseun Oladeji Olaniyi, Oyewole Dokun, Udoka Felista Eze, O. C. Nwokonkwo
Abstract
This paper extensively explores the global impact of COVID-19 on mortality, society, and health, emphasizing the pressing need for accurate prediction models to navigate the ever-changing virus strains. Deep learning models play a central role in numerous phases of COVID-19 prediction, as well as diagnosis, severity assessment, and therapeutic expansion. Challenges in virus diagnosis, especially the time-consuming RT-PCR test, are addressed by leveraging AI and deep learning algorithms to predict transmission patterns and identify high-risk patients using diverse data sources like X-rays, CT scans, clinical records, and lab results. The assessment of disease severity, spanning from mild symptoms to serious conditions like renal failure and heart damage, is a key focus. Deep learning frameworks offer promising alternatives, reducing resource-intensive testing while enhancing detection and in-depth analysis of complex medical data. The paper provides a systematic taxonomy categorizing research papers by research questions, offering a comprehensive overview of COVID-19 prediction research trends and gaps, with suggestions for exploring diverse data modalities, improving model interpretability through XAI, and integrating parallel processing. In summary, this paper offers valuable insights into the COVID-19 prediction landscape, serving as a vital resource for academics and researchers.