How AI could help in the fight against COVID-19
From developing drug treatments to predicting the next hotspot, artificial intelligence may help researchers, healthcare workers, and everyday people offset the impact of the coronavirus.
As the worldwide fight against coronavirus COVID-19 continues, companies and governments around the world are pulling out all the stops in an effort to stave off the pandemic’s worst impacts. One tool in that toolbox that might prove particularly useful is artificial intelligence (AI). Even though AI has been around since the 1960s, it’s only been in the past few years that its adoption outside of science labs and research institutions has really taken off.
Chatbots, RPA, and telemedicine
Perhaps the most common application of AI people have come into contact with today are virtual assistants like Apple’s Siri and Amazon’s Alexa, which rely on natural language processing (NLP) algorithms to understand human speech.
This ability could prove particularly useful in relieving the administrative burden on over-stretched and under-resourced healthcare staff by deploying AI-based chatbots to prescreen patients for COVID-19 symptoms, said Tanya Berger-Wolf, director of the Translational Data Analytics Institute at The Ohio State University (OSU). These screenings could be done onsite or remotely via telemedicine for isolated populations or for at-risk individuals who do not want to, or cannot, leave their homes.
SEE:Coronavirus: Critical IT policies and tools every business needs(TechRepublic Premium)
Suki, an AI-based, voice-enabled digital clinical assistant startup that helps doctors create medical notes and do other things like retrieve patient information, has re-engineered its offering with COVID-19-specific templates to speed up the note-taking process. The startup is also looking at deploying its technology for use in telemedicine, an area of technology that is taking off right now.
Robotic process automation (RPA), a form of AI that can handle complex routine tasks, could then be deployed to do things like process all of the clinical and insurance forms that a healthcare facility would need to move the patient-evaluation process forward, said Berger-Wolf.
“The use of natural language processing is going to be critical,” she said.
Predicting COVID-19 hotspots and how to respond
Another area where AI can be useful is in predicting future events. OSU’s public health department is working on an AI algorithm that uses “what-if” scenarios, like applying different dates for lifting stay-at-home orders, to figure out when COVID-19 will crest in Columbus, OH, where OSU’s main campus is located.
“What comes out of that is a much more accurate focus forecast,” said Berger-Wolf. “If we start easing them, what will happen? We can start modeling it. And the more data we have, the better we can model it. We can start understanding the correlational relationship between the type of person and the circumstances of their life, for example, and then the outcome of infection.”
Harry Glaser, the co-founder and CEO of data analytics firm Periscope Data, a SiSense company, believes that AI could predict the next COVID-19 hotspot, provided the high-quality data these algorithms need is available.
“The technology could absolutely be brought to bear to predict the next hotspot,” he said. “If you had the level of data and tracking that you have right now in a place like China, you would be able to have high confidence in a prediction about where it’s going to go and where the hotspots are going to be next.”
According to the research paper, A dynamic neural network model for predicting risk of Zika in real time, published on BMC Medicine, this was done successfully during the Zika outbreak in 2015. “We show that the model can accurately predict the geographic expansion of Zika in the Americas with the overall average accuracy remaining above 85% even for prediction windows of up to 12 weeks,” the paper said.
“Early on, researchers calculated the potential spread rate of a single UK citizen who had traveled in China, then back to the UK, and then in Italy,” said David Bloch, data science evangelist at Domino Data. “Other useful and available data sources include location data, pathing data, and ‘active carrier’ data on infected individuals.”
The problem with developing such models in a country like the US, which places a high value on individual liberties and privacy, is you need very granular data on people’s movements and who they interact with, said Glaser. Countries like China, where this is not an issue, have already deployed cell phone apps to conduct contract tracing of people infected with COVID-19. These apps are being deployed in other Asian countries and are being explored by Western governments, as well.
Given the right data, you could also develop models to predict the severity of an outbreak based, potentially, on factors like population density and age, for example.
“Maybe the reason the death rate appeared to be higher in Italy is because the health system got overloaded, and so people didn’t get life-saving care,” said Glaser. “And maybe the reason this virality is so high in New York is because it’s so dense. Those are plausible theories. You pop them into a model as variables and understand really quickly the different environmental factors and systemic factors that affect these characteristics of the virus. And, then, that could tell you what to do. Maybe a less dense area doesn’t need to go into lockdown quite as severely as New York does, or quite as early as New York does, as an example.”
Diagnostics is also an area where AI can prove very helpful. Medical imaging companies are working on AI-based algorithms that can diagnose COVID-19 from CT scans and X-rays.
ElectifAi, a developer of machine learning products for multiple industries, is one of many companies that are developing ways to detect and track COVID-19 in the lungs of affected patients by applying machine learning, a more advanced form of AI, to CT scans. The company said it is “working around the clock” collecting anonymized CT scans from healthcare experts around the world to train and optimize its models.
ElectifAi is asking medical professionals who are interested in their project to get in contact with them directly.
“This is an important step for understanding the underlying impact of the illness, its progress and development, and provides medical professionals with another tool to relieve the strain they are facing on a daily basis,” said Edward Scott, ElectrifAi’s CEO.
Drug development could be the area that has the most impact on negating the worst effects of the pandemic. AI has been used in drug development for years, and researchers are turning to these models today to identify potential treatments for COVID-19.
“Natural language processing permits researchers to leverage machine learning to search through reams of past academic and media papers at speeds that are hundreds of times faster than they could otherwise work,” said Block. “By doing so, they create a culled list of relevant papers that saves researchers time as they find answers to pressing questions.”
On March 25, the Singapore-based Gero, which has an AI-driven drug discovery platform, announced that the company:
… used its AI platform to identify the potential anti-COVID-19 agents that have been previously tested on humans. Six of them have been approved, three were withdrawn, and the other nine have been already tested in clinical trials. The emergency of the situation, as well as the legal and regulatory status of these agents, make it possible to start immediate clinical trials for most of the suggested drugs.
The company used its AI drug discovery platform to identify molecules with potential effects on the coronavirus replication. The fact that this time the potential treatments were found among the existing drugs marks a significant improvement over previous efforts to use AI to predict molecules active against COVID-19. The discovery makes it possible to start clinical trials in a matter of weeks.
Gero’s press release includes more details about the drugs found to be potentially effective and where those drugs are approved.
This is just one example. A quick Google search reveals that AI is being used in corporate research labs, national institutes of health, and universities all around the globe to identify potential treatments and cures.
SEE:Coronavirus vaccines: The drugs in development to treat COVID-19(CNET)
“That is prime use of AI and, right now, several companies are doing it, including Google’s DeepMind,” said Berger-Wolf.
AI can further speed the drug-development process by modeling the efficacy of these drugs prior to clinical trials, said Bloch.
“Simulations of real tests narrow the field of actual required tests to be performed,” he said. “The result is faster and more efficient research when time is of the essence. Being able to readily reproduce prior experiments and make adjustments for COVID-19 greatly reduces the research time involved.”
Image: metamorworks, Getty Images/iStockphoto