AI may open up promising new ways to combat Covid-19, but remain cautious
In December, a Canadian start-up was one of the first to identify the threat of an epidemic spreading in Hubei province and inform its customers , including business and government agencies in several countries. How? Through an AI algorithm that analysed health data to identify emerging trends before anyone else.
Four months later, the epidemic has become global and no one is sure how the world will emerge. It’s clear that the early diagnosis didn’t avoid the crisis: proof that we have to be careful of technological “solutionism”.
However, AI could prove useful at the end of the crisis, provided that it is seen as an auxiliary and not fix-all, and that the health crisis doesn’t undermine the protection of personal data.
Understanding the virus better
Medically, AI can help us better understand how the virus works and identify treatments.
DeepMind, the Google subsidiary specialising in AI, has announced it is sequencing the protein of the SARS-CoV-2 virus through its AlphaFold tool, which makes it possible to model protein in 3D. “We hope to contribute to the scientific effort using the latest version of our AlphaFold system by releasing structure predictions of several under-studied proteins associated with SARS-CoV-2, the virus that causes COVID-19,” explains a DeepMind press release. DeepMind has shared the results with the scientific community. But the company warns against being overly confident because the results have yet to be verified by peers and are only preliminary: the experiences necessary to identify these structures can take months and sometimes lead to a dead end.
In France, the National Institute of Health and Medical Research (INSERM) announced funding for a similar research project using artificial intelligence to better understand the biochemistry of virus proteins and test the effect of possible inhibitors on the samples obtained.
Screening and diagnostics
In a context where hospitals and healthcare professionals are overloaded and the number of tests is insufficient, companies are developing artificial intelligence solutions to help the screening process.
Simple diagnostic tools based on chatbots already exist , but they don’t do much beyond the government’s self-diagnosis tools already available online.
Other more ambitious uses are based on computer vision and image processing with AI software (as is already the case in ophthalmology or dermatology , with very convincing results).
A Chinese company, Infervision, announced it had developed computer vision software capable of analysing chest scanners for signs related to the virus, including lung complications.
According to Infervision’s founder, interviewed in The Lancet medical journal , diagnostic aids of this type could also reduce infection among healthcare workers: “An article in JAMA states that human-to-human hospital-associated transmission accounted for 41% of all cases in a study of patients at Zhongnan Hospital of Wuhan University. We also know that more than 1000 hospital staff in the city of Wuhan have been confirmed infected.”
However, the reliability of the diagnoses carried out to date has not yet been proven. What’s more, these tools can only be used in cases where the virus has been declared, not in asymptomatic patients or mild cases.
The search for treatments
The massive processing capabilities of AI software and their ability to detect meaningful patterns, which are undetectable on a smaller scale, can be very useful in finding treatments.
According to the magazine Wired , the White House has launched an initiative with academics and tech companies to enable AI researchers to access a vast body of scientific information related to the Coronavirus. “By cross-referencing papers and searching for patterns, AI algorithms might help discover new possible treatments or factors that make the virus worse for some patients.” states Wired.
Several companies are using artificial intelligence tools, combined with genomics and precision medicine techniques to try to identify treatments.
Some techniques take advantage of the massive computing power and ability of AI to perform a very large number of calculations to look for possible drugs: this is the case with “generative design” algorithms, which produce a wide range of results that can be sorted to select only the most pertinent.
Another method is to combine the massive amount of scientific information with genomic techniques to identify existing drugs that can help treat the virus.
Pharmanext in France and BenevolentAI in the UK have announced they have identified several potential drugs using AI.
While it is clear that AI is a valuable tool to aid medical research, caution must be taken when it comes to press coverage, in medical AI as elsewhere. These initiatives are recent and their results haven’t been tested over time nor validated by the scientific community.
Here, as elsewhere, the warnings posted by Inserm , which is very active in AI, remain valid: “In spite of the enormous calculation capacities offered by current computers, no existing application can claim to be genuinely intelligent: for that it needs to be multitasking and able to react correctly in unforeseeable and non-preprogrammed situations. There is a long way to go yet.”