Deep Learning applications for COVID-19

Abstract

This survey explores how Deep Learning has battled the COVID-19 pandemic and provides directions for future research on COVID-19. We cover Deep Learning applications in Natural Language Processing, Computer Vision, Life Sciences, and Epidemiology. We describe how each of these applications vary with the availability of big data and how learning tasks are constructed. We begin by evaluating the current state of Deep Learning and conclude with key limitations of Deep Learning for COVID-19 applications. These limitations include Interpretability, Generalization Metrics, Learning from Limited Labeled Data, and Data Privacy. Natural Language Processing applications include mining COVID-19 research for Information Retrieval and Question Answering, as well as Misinformation Detection, and Public Sentiment Analysis. Computer Vision applications cover Medical Image Analysis, Ambient Intelligence, and Vision-based Robotics. Within Life Sciences, our survey looks at how Deep Learning can be applied to Precision Diagnostics, Protein Structure Prediction, and Drug Repurposing. Deep Learning has additionally been utilized in Spread Forecasting for Epidemiology. Our literature review has found many examples of Deep Learning systems to fight COVID-19. We hope that this survey will help accelerate the use of Deep Learning for COVID-19 research.

Keywords: COVID-19; Computer Vision; Deep Learning applications; Epidemiology; Life Sciences; Natural Language Processing.

Fig. 1

Organization of Artificial Intelligence COVID-19 Applications, comparison with other literature surveys

Fig. 2

What tasks has NLP conquered? A quick overview of the number of examples, tasks, and domains contained in the GLUE benchmark (Image taken from Wang et al. [29])

Fig. 3

Interface for human labeling TREC-COVID documents (Image taken from Voorhees et al. [43])

Fig. 4

CO-Search System Architecture (Image taken from Esteva et al. [38])

Fig. 5

User Interface for the CAiRE-COVID Literature Mining system consisting of Extractive Summary, Abstractive Summary, and most relevant documents (Image taken from Su et al. [39])

Fig. 6

A Knowledge Graph organization of our survey on Deep Learning to fight COVID-19. Here every relation is “A contains B”

Fig. 7

BenevolentAI Knowledge Graph used to suggest baricitinib as a treatment for COVID-19 (Image taken from Richardson et al. [8])

Fig. 8

Meta-data on the count of Entities in CKG and Relation information (Image taken from Wise et al. [60])

Fig. 9

Examples of Misinformation Labels (Image taken from Hossain et al. [69])

Fig. 10

COVID-19 claim examples about COVID-19 and corresponding evidence retrieved (Image taken from Wadden et al. [71])

Fig. 11

AUC and Accuracy performance gains from ConVIRT (Image taken from Zhang et al. [97])

Fig. 12

An illustration of Ambient intelligence of daily living spaces in an elderly home equipped with one ambient sensor (Image taken from Haque et al. [10])

Fig. 13

Rough overview of RT-PCR amplifications (Image taken from Lopez-Rincon et al. [119])

Fig. 14

AI for Precision Medicine (Image taken from Zhou et al. [122])

Fig. 15

Illustrations of how different Neural Network architectures have been applied to Protein Structural Modeling (Image taken from Gao et al. [127])

Fig. 16

Illustration of the use and construction of CoV-KGE for drug repurposing (Image taken from Zeng et al. [61])

Fig. 17

Hi-COVIDNet, using country-level and continent-level encoders to predict imported COVID-19 cases from travel (Image taken from Kim et al. [133])

Fig. 18

SEIR differential equations (Image taken from Dandekar and Barbastathis [135])

Fig. 19

The Neural Network takes an input vector of the Susceptible, Infected, Recovered, and Quarantined Population (estimated from the previous time step) to model the quarantine strength (Image taken from Dandekar and Barabastathis [135])

Fig. 20

Further differentiation within populations in the SIR model (Image taken from Arik et al. [136])

Fig. 21

Graph-structured SIR models (Image taken from Meirom et al. [137])

Fig. 22

Privacy-preserving techniques explored in applications of Ambient Intelligence (Image taken from Haque et al. [10])