In this post, we’re collecting articles that detail how AI, machine learning, and data science are being used in response to the coronavirus. So far it covers predictive and monitoring systems, AI-driven diagnostics, and efforts in the data science community. We hope it will provide a comprehensive, grounded look at exactly where AI technology is being used, and how it is making a difference. In time, we hope it will also provide an overview of how AI technology was employed to understand, predict, and fight the spread of COVID-19.
This post will be regularly updated as new developments are reported.
AI Platform BlueDot Detects COVID-19 on December 30th, 2019
Not long after COVID-19 was officially announced, news sites reported that the AI-powered BlueDot platform had detected the beginnings of the virus on December 30th, 2019. BlueDot is designed to track, locate, and conceptualize the spread of infectious diseases. It does this via machine learning algorithms which analyze data from thousands of sources, including public health organization, population demographics, and even livestock health reports. It updates every 15 minutes, and its data is reviewed by a team of health experts and programmers.
The CDC Looks to Machine Learning to Forecast the Spread of COVID-19
The US Centers for Disease Control and Prevention has enlisted the help of dozens of teams in an effort to forecast the spread of the coronavirus. Among these is a group from Carnegie Mellon University that utilizes machine learning algorithms to read data streams and develop real-time predictions from them, supported by human-based reviews.
South Korea Utilizes Big Data and AI to Stay Ahead
According to this news report, South Korea is making use of big data analysis and AI-powered advance warning systems to keep informed regarding the spread of COVID-19. The government-run analytical system contains information on citizens, foreign nationals, hospitals, government organizations and financial services to develop maps of locations visited by the infected so it can warn people of potential clusters of the virus. The government has also implemented AI-based regulation to balance the distribution and supply of masks and other medical items.
The analytical systems shared below are open to the public and designed to give people around the world up to date information on the trajectory of the virus, reported cases, and growth rate, both globally and by country.
COVID-19 Dashboards Provide Global Visualizations on the Hour
This collection of automated visualizations updates every hour with statistics and predictions regarding the growth rate of the virus, global exploratory data analytics, confirmed cases in the US, and estimated mortality rates. It runs primarily on data from the John Hopkins COVID-19 Data Repository.
Global Analytics Dashboard
In order to give people up to date information about coronavirus at both a global and country-wide scale, Anodot has launched a public service that includes an analytics dashboard based on machine learning. This dashboard tracks reported cases of COVID-19 using data from both the John Hopkins University and The Center for Systems Science and Engineering.
ClosedLoop.ai Releases AI-based Identification Model
ClosedLoop’s AI model aims to identify the people most vulnerable to complications caused by COVID-19 by predicting the likelihood of a hospital stay due to respiratory infections, such as influenza or pneumonia. It was built and tested using medical claims data from more than 2 million elderly and disabled individuals. Their open-source AI-based toolkit – the CV19 Vulnerability Index – can be found here.
Microsoft Mobilizes AI for Health Initiative to focus on Coronavirus
Microsoft announced the launch of AI for Health on January 29, 2020, and on April 9th has decided to mobilize that initiative to support research into COVID-19. Along with working to support remote education and businesses, Microsoft is engaging in partnerships that include providing researchers with computing resources, working with the Institute of Health Metrics and Evaluation, working on a new dashboard with the Washington State Department of Health, and developing chatbots to help avoid overloading in hospitals. Read more about each of these in detail at their announcement blogpost here.
ElementAI releases Free Platform to Assist Response Efforts
ElementAI have developed a platform configured to the COVID-19 Open Research Dataset (CORD-19), a dataset of papers and articles related to COVID-19 research. Their platform utilizes a semantic search model, which allows users to make queries in natural language, then returns any articles that are semantically similar. The hope is that by creating drawing together similarities, researchers may uncover potential patterns across research papers and datasets.
AI systems have been improving medical diagnostics in a variety of ways, including predicting viral structures, reading CT scan data, and predicting protein structures for the development of therapeutics.
Baidu’s Linearfold Algorithm Improves Virus Analysis Times by 120 Times
To support research into improved diagnosis of COVID-19, Baidu made its Linearfold algorithm available to scientific and medical teams working on diagnostic systems. From their statement: “The Linearfold algorithm, published in partnership with Oregon State University and the University of Rochester in 2019, is significantly faster than traditional RNA folding algorithms at predicting a virus’s secondary RNA structure. Analyzing the secondary structural changes between homologous RNA virus sequences (such as bats and humans) can provide scientists with further insight into how viruses spread across species. Due to the recent outbreak, Baidu AI scientists have used this algorithm to predict the secondary structure prediction for the Covid-19 RNA sequence, reducing overall analysis time from 55 minutes to 27 seconds, meaning it is 120 times faster.”
Research Paper Details Deep Learning System for Screening the Coronavirus
This research paper details how deep learning systems can be used to determine between illnesses and make probability ratings for potential cases of COVID-19. The model introduced in the paper is trained on CT scan data from COVID-19 patients, influenza patients, and healthy people from hospitals in Wuhan, China.
Alibaba Says its AI Model Identifies COVID-19 Infections with 96% Accuracy
Alibaba’s research institute, Damo Academy, says their AI-powered diagnostic system can detect coronavirus cases with an accuracy rate of up to 96%. The diagnostic system identifies the differences in CT scans between coronavirus and ordinary viral pneumonia. It is trained on sample data from more than 5,000 confirmed cases. According to Damo Academy, the algorithm is capable of running scans in 20 seconds, and also includes the latest treatment guidelines and recently published research. Alibaba says the system will be adopted across more than 100 hospitals in the provinces of Hubei, Guangdong, and Anhui.
Preprint Paper Demonstrates Proof-of-Principle For Using AI in Diagnostics
This paper, by experts from universities, hospitals, and laboratories across China, details another deep learning algorithm for use in diagnostics through CT scans. It is trained on 1,119 images collected from pathogen-confirmed COVID-19 cases and those diagnosed with typical viral pneumonia. The algorithm achieved a total accuracy of 89% for internal validation, and 79% on the external testing dataset. *This paper has yet to be peer-reviewed as of March 23rd, 2020.
Deepmind Releases Computation Predictions of Protein Structures of COVID-19
Using the latest version of their Alphafold system, Deepmind has released structure predictions of several proteins associated with SARS-CoV-2, the virus that causes COVID-19. The AlphaFold system uses deep learning algorithms to predict protein structures and generate 3D models of proteins at very high levels of accuracy. Though they have yet to be experimentally verified, Deepmind hopes the predictions will help scientists better understand how the virus functions and develop future therapeutics.
University of Trento: Italian COVID-19 Lung Ultrasound Project
Abbreviated as ICLUS, the goal of this project is to develop an automatic diagnosis, monitoring, and reporting system of coronavirus patients through AI analytics applied to ultrasound images. According to their announcement, “ICLUS defines a specific measurement protocol for optimal settings of ultrasound instruments, the 14 acquisition points, and a 4-level scoring system to establish the severity of a patient’s condition.” The group behind the project is made up of professors and researchers in a variety of different specialized areas including ultrasound, signal processing and recognition, and deep visual learning.
Delft Imaging and Thirona Give Away AI-based Diagnosis Tool
CAD4COVID is a cloud-based system that uses deep learning to analyze chest X-rays for COVID-19 patients. The developers hope the tool will speed up the processing of patients and reduce pressure on healthcare systems, particularly in areas where reliable CT scans are not available. Find out more about the CD4COVID tool here.
Efforts In the Data Science Community:
The MIT-IBM Watson AI Lab funds 10 research projects addressing the pandemic
Believing that AI technology can play a decisive role in responding to the COVID-19 pandemic, the MIT-IBM Watson AI Lab is funding 10 research projects focused on AI technology that addresses health and the economy. These research projects include early detection, protein design, face mask materials, and privacy as it relates to contact tracing. The details of all 10 of these research projects can be found here.
COVID Moonshot Project by PostEra
PostEra specializes in medicinal chemistry as powered by machine learning. Their COVID Moonshot project is a way of increasing the number of inhibitor designs that could potentially result in effective anti-COVID drugs. They are asking for interested chemists to design and create compounds that can be tested, and a community is building around the project with its own dedicated discussion forum.
COVID-19 EHR Dream Challenge
The Electronic Health Record (EHR) Dream Challenge is a challenge for data scientists to build a model that answers the question: “Of patients who have at least one clinical encounter/visit at UW (University of Washington) Medicine and who were tested for COVD-19, can we predict who is positive?” Participants will have access to synthetic data in the cloud for testing the validity of their models, with the overall goal of incorporating machine learning and predictive algorithms into clinical care.
Allen Institute for AI releases the CoViz Search Tool
CoViz is designed to give researchers a more comprehensive picture of the information contained in the COVID-19 Open Research Dataset. In particular, it helps researchers explore associations between concepts through a sciBERT model that was trained on a corpus of research papers and fine-tuned on several biomedical entity recognition tasks. The Allen Institute of AI plans to continue development of CoViz to make it even better at handling scientific language. You can explore CoViz yourself here.
Stanford University Hosts Virtual Conference: “COVID-19 and AI”
The Stanford Institute of Human-Centered AI is holding a virtual conference on Wednesday, April 1st, centered on areas in which AI intersects with ongoing research on the Coronavirus. The conference brings together experts from Stanford as well as researchers and engineers from across the globe. Topics will be covered through presentations and interactive sessions, and include AI applications in diagnostics, information and disinformation, and the impact of pandemics on economies. RSVP and agenda details can be found here.
The Covid-19 Global Hackathon
With support from tech companies and platforms including Facebook, Microsoft, TikTok, Twitter, and more, this hackathon aims to bring together developers to build software solutions that tackle challenges related to the coronavirus. The organizers also hope that the event will create an online space for continued development around themes including health, community, vulnerable populations, and education. The winners will be announced as of April 10th, and results will be available here.
Covid Act Now Released to Public
Covidactnow.org is a tool built to enable political leaders to make decisions regarding response to the Coronavirus. It brings together available data to create a map and model designed to help leaders determine the impact in their regions as well as the potential pressure on healthcare systems. The tool was created by a team of data scientists and engineers working together with epidemiologists and public health officials.
COVID-19: A Data Science Perspective
Jeremy Howard and Rachael Thomas of fast.ai have written a long, informative piece analyzing COVID-19, covering what it means for the healthcare system, how the coronavirus is different from the flu, flattening the curve, and more. Their report is a comprehensive look at what data can tell us specifically in terms of the growth of the virus, and what it has told us about similar (if much smaller) events in the past. The post is available in more than 15 languages.
COVID-19 Datasets Shared Globally
Datasets from every country with confirmed cases of the coronavirus have been compiled in this repository, ranging from map data and country specific cases to fatality rates and gender distribution. Among these datasets, the White House Office of Science and Technology Policy (with representatives from Microsoft, the Allen Institute for AI, the Chan Zuckerberg Initiative, and others) recently released a dataset containing 29,000 scientific articles related to the coronavirus. Their hope is for AI researchers to analyze the data on behalf of the scientific community.
AI Technology and Robotics:
Facebook uses AI to detect COVID-19 misinformation
The Facebook Artificial Intelligence group have put consistent work into AI systems that can respond to the spread of misinformation regarding COVID-19. Some of these systems are designed to help fact checkers by detecting near-exact duplicates of content flagged as potential misinformation, while others are vision models specifically trained to detect potentially exploitative images in ads and online marketplaces.
Frances uses AI to monitor mask wearing on public transport
According to The Verge, France is implementing AI technology into security cameras to confirm that people are wearing face-masks while using public transport. The data is not intended to enforce mask wearing rules, but rate to publish statistics related to the number of people wearing masks in the city. The technology is currently undergoing testing as part of a three month trial started in May.
Robots Used in Korea to Disinfect, Send Social Distance Warnings, and Check Temperatures
According to VentureBeat, self-driving robots are being used at SK Telecom, South Korea’s largest mobile operator. The robot is equipped with cameras and an LED screen, which it uses to remind people of social distancing rules. At the same time, the robot can also check people’s temperatures, dispense hand sanitizer, and disinfect the floor. The developers, Omron Electronics Korea, have also made sure to add a function that hides faces to protect privacy.
Spain Automates Virus Testing with Robotics
The Spanish government recently announced a plan to increase daily testing through the implementation of robotic testers. The automation will reportedly increase daily testing rates from 20,000 to 80,000 tests per day. Exactly how the robotic automation will work has yet to be announced as of March 23rd.
China Uses Disinfection Robots
Self-driving disinfection robots have been employed in hospitals in China to help mitigate the spread of COVID-19. The robots, developed by UVD Robots, scan and kill bacteria autonomously through the use of ultraviolet light.
If you are unsure what to do or who to contact with questions or concerns regarding COVID-19, the WHO has a dedicated Coronavirus page that includes how to stay safe, research around the world, and constantly updating situational updates.
The team at Lionbridge AI is currently working remote, and we’ll be following government guidelines and advice as the situation in Japan and the world evolves over the coming weeks and months. However, we expect business as usual, so don’t hesitate to get in touch if you have any questions, concerns, or are in need of assistance.
Remember to avoid large crowds where possible, wash your hands, avoid touching your face, and above all, stay safe.