Artificial Intelligence on the Frontlines Against COVID-19

COVID-19 Dashboard at Johns Hopkins University (September 14, 2020)

While many may not recognize artificial intelligence (AI) on the ground during the unfolding COVID-19 pandemic, it is increasingly being implemented on the front lines. It is no longer the case that AI is restricted to academic circles and research institutions, but it is one of the most effective forms of defense and prevention.

One of its most important contributions comes from its ability to effectively screen, track, analyze and make predictions for both current and future patients. AI can be utilized for contact tracing, and it identifies high-risk areas within the nation and where the virus could be headed.

By processing big data, AI can accurately pinpoint hot zones, helping officials decide which schools and workplaces should remain open or shut down. Targeted isolation strategies such as this could help blunten the impact of quarantine measures. In China, where the outbreak began, the technology company Baidu developed a no-contact infrared sensor system capable of identifying individuals with fevers. The system is effective in crowds and is being implemented in railway stations within the nation.

The COVID-19 pandemic has resulted in the acceleration of AI implementation into hospitals around the globe, specifically being used by healthcare professionals and experts to identify ‘red flags,’ as well as speed up decision making. By being paired with current medical imaging technologies such as CT and MRI, it can assist in the development of more efficient diagnosis methods.

Tampa General Hospital in Florida has an AI system present at the entrance of the building, and it is able to identify potential COVID-19 symptoms to prevent individuals from coming in contact with existing patients. This system relies on cameras capable of conducting a facial thermal scan, which can then pick up symptoms like sweat and discoloration.

Back in China, the country is taking AI even deeper into the hospital system with an AI-driven scan interpreter at Zhongan Hospital. This allows COVID-19 to be identified in a patient even if there are no radiologists available. Taking it a step further, Wuhan Wuchang Hospital developed a section staffed mostly by robots. Through the use of these intelligent robots, vital signs can be monitored, medicine and food delivered to patients, and the system results in less exposure to the virus for physicians.

Treatments, Vaccines and Prevention

AI is an effective tool for drug research, capable of analyzing data on COVID-19 and helping with drug delivery design and development.

Private companies like IBM are playing a big role in drug research by releasing technologies such as the Corpus Conversion Service and Corpus Processing Service, both of which were already present in existing industries.

The Corpus Conversion Service is able to take in 100,000 PDF pages per day on a single server. It then relies on advanced machine learning models to extract highly accurate content. This specific technology is being used on COVID-19 and coronavirus-related PDFs, as well as being combined with other important databases.

The Corpus Processing Service integrates data from databases and publications in order to gain insight into important points, such as current drugs, their outcomes and new risk-factors.

AI is crucial for speeding up the drug testing phase, which traditionally takes a significant amount of time with current methods. When the drug finally does make it to clinical trials, AI can be called on once again to quicken the process.

Looking Outside Traditional AI Models for COVID-19

One of the major challenges surrounding the use of AI to combat the COVID-19 pandemic in the early days had to do with the speed at which the virus spreads. Looking just at case numbers, the global count went from less than 100,000 in early March to over 13 million by July, including 580,000 deaths. Besides the immediate health repercussions, there were rapidly evolving economic and supply chain crises.

Since traditional machine learning and advanced analytical models take around three to four months from start to finish, there needed to be a shift to models that require significantly less set-up time.

Agent-based modeling relies on intelligent agents, which are computer applications that aim for designed objectives by autonomously sensing and responding to the environment. These intelligent agents can include technology like softbots, taskbots and personal agents, all of which include one or more of the following features: Autonomous, Adaptive/Learning, Social, Mobile, Goal-oriented, Communicative and Intelligent.

Intelligent agents are used to simulate, through their actions and interactions in an environment, the complex dynamics of a system. This allows different parameters and rules to be set for individual agents, and the biggest benefit is the ability to model individual styles and attributes.

In regard to the COVID-19 pandemic, agent-based modeling was used to analyze human behavior in the beginning, such as whether or not individuals were complying with the stay-at-home orders. This modeling type relied on daily mobility data from different parts of the U.S., and it played a key role in learning about the differences in how and where people moved.

System dynamic modeling can establish a better understanding of the behavior in a system by using a model to represent the relationships in that system. This allows simulations to be run, which result in enhanced forecasting and planning.

This modeling type was used to bring together decisions and disruptions in one place for COVID-19, such as the progression of the disease, government initiatives and the behavior of populations.

Agent-based modeling and system dynamic modeling fall under the category of minimum viable AI model, or MVAIM. One of the best examples of this is A SEIRD (Susceptible-Exposed-Infected-Death), which is a disease progression model that estimates the COVID-19 risk to certain population groups. This model was heavily relied on during the early outbreak of the pandemic, carrying out tasks like estimating hospitalizations and ventilators.

The model was used to track COVID-19 progression in all 50 U.S. states, eventually becoming more sophisticated and expanded into all counties within the country. This MVAIM was able to be developed in one week, and it took just another to be tested, validated and deployed.

Because human behavior is unpredictable during a pandemic, the importance of this model rises, making it a valuable tool for future outbreaks.

Many experts believe that if it wasn’t for the lack of resources dedicated to AI development in many nations, such as the United States, the technology could have seen the virus coming far in advance. It is very possible that if AI had been widely utilized before or right at the start of the outbreak, there would be far less case numbers, death and damage to the economy. It is not too late to learn from this situation and better prepare society for future outbreaks, whether it be another form of COVID-19 or some unknown virus.

While the world fights its way out of this pandemic, expect there to be more attention and funds dedicated to research and development in these areas. Artificial intelligence has already proven to be one of the best tools available to humans throughout our entire history, we just need to turn the corner on it.