Deep Learning Techniques for COVID-19 Diagnosis and Prognosis Based on Radiological Imaging

This literature review summarizes the current deep learning methods developed by the medical imaging AI research community that have been focused on resolving lung imaging problems related to coronavirus disease 2019 (COVID-19). COVID-19 shares many of the same imaging characteristics as other common forms of bacterial and viral pneumonia. Differentiating COVID-19 from other common pulmonary infections is a non-trivial task. To help offset what commonly requires hours of tedious manual annotation, several innovative solutions have been published to help healthcare providers during the COVID-19 pandemic. However, the absence of a comprehensive survey on the subject makes it challenging to ascertain which approaches are promising and therefore deserve further investigation. In this survey, we present an in-depth review of deep learning techniques that have recently been applied to the task of discovering the diagnosis and prognosis of COVID-19 patients. We categorize existing approaches based on features such as dimensionality of radiological imaging, system purpose, and used deep learning techniques, underlying core issues, and challenges. We also address the merits and shortcomings of various approaches, and finally we discuss future directions for this research.

Automated summary

This literature review provides an overview of current deep learning methods used in medical imaging AI research to tackle lung imaging problems related to COVID-19. Due to similar imaging characteristics, distinguishing COVID-19 from other pulmonary infections is challenging. Several innovative solutions have been proposed to assist healthcare providers, but the lack of a comprehensive survey makes it difficult to determine promising approaches. This survey reviews recent deep learning techniques for diagnosing and predicting outcomes of COVID-19 patients, categorizing them based on imaging features, system purpose, deep learning techniques, core issues, and challenges.