5 Ways Healthcare Organizations and Researchers are Fighting Covid-19 with AI.
On Dec. 30 of last year, Healthmap, an artificial intelligence (AI)-driven data mining system that scans disparate data sources for signs of disease outbreaks, spotted unusual activity about a new type of pneumonia cropping up in China. One day later, AI outbreak risk software solution BlueDot began sounding a similar alarm after scanning thousands of Chinese news reports, along with airline and animal disease network data, with its machine learning algorithms.
Both solutions used AI to help alert the international community about what eventually would become the COVID-19 pandemic. Since then, AI-based techniques have been applied to virtually every aspect of the outbreak, from prediction to diagnostics and treatment.
And although experts say AI’s success in fighting COVID-19 has been middling at best so far, this is changing quickly, according to Kai-Fu Lee, CEO of Chinese VC firm Sinovation Ventures. “Before COVID-19 struck, we did not understand the importance of these areas and act accordingly, and, crucially as far as AI is concerned, we did not have the data to deliver the solutions,” he says in Wired.
But he says the pandemic soon led to an explosion in the availability of machine readable healthcare data, led by the COVID-19 Open Research Dataset Challenge, known as CORD-19. This, combined with transparent data sharing and cooperation among tens of thousands of scientists – all working “with a transparency and at speeds we’ve never seen before” – has changed the game entirely.
How AI is leading the fight against COVID-19
There’s no doubt by now that the pandemic has accelerated the ever-increasing union between AI and healthcare in general. But it’s also helping shape the fight against COVID-19 in particular. In a research paper published by the medical journal Diabetes & Metabolic Syndrome: Clinical Research & Reviews, Vaishya et al. say there are seven significant ways AI can be applied to COVID-19 including early detection, diagnosis, contact tracing and prevention. We’ll discuss a few of these applications, and more, below.
1. Finding or predicting new outbreaks
The growing importance of AI, big data analytics and machine learning when it comes to outbreak detection and prediction can’t be overstated. We’re now seeing technology companies and AI researchers team up on automated tracking systems, for example, that mine huge tracts of data for warning signs. And researchers at California’s Chan Zuckerberg Biohub are building AI models that attempt to quantify undetected COVID-19 infections within a certain region, to help forecast viral spread in the future. Similarly, other organizations such as Closedloop.ai have developed an AI-based vulnerability index, meant to predict and identify those most at risk of the worst complications from COVID-19.
2. Speeding up (and improving) diagnostics
AI and machine learning are improving diagnostics in several ways, but none more effectively than in the analysis and annotation of medical images. Some companies are now using AI technology that find and mark anomalies such as ground-glass opacities within chest X-rays, along with quantifying ongoing infection volumes to assist with patient monitoring and management. CapeStart offers services to train AI systems to detect anomalies. Similarly, researchers at UC San Diego Health recently came up with a new way to diagnose pneumonia (a well-known condition associated with COVID-19) sooner, which they believe will help when assessing coronavirus patients with complications.
3. Speeding up (and improving) medical research
We already mentioned the explosion in publicly-available data that has accompanied the pandemic, including CORD-19, hosted by Kaggle and consisting of more than 128,000 machine readable research papers and other materials on coronaviruses. Not long after its release, Amazon Web Services (AWS) released CORD-19 Search, an AWS-based search engine for the CORD-19 database. And in Russia, the largest CT scan database in the world for COVID-19 purposes has been assembled at Moscow’s Diagnostics and Telemedicine Centre, containing more than 1,000 sets of chest CT scans. These and other datasets and tools are helping to further speed up research, while also serving as accurate training data for machine learning models.
4. Finding treatments and new drugs
Pharmaceutical companies often use NLP technology for drug discovery, drug repurposing, the identification of drug targets, and to enhance drug safety and efficacy, and the same techniques are playing a role in accelerating the discovery of drugs to help treat COVID-19. UK-based BenevolentAI, for example, recently used its drug discovery platform to scan thousands of approved drugs for possible benefits for treating COVID-19. Within days it had identified the rheumatoid arthritis drug Barictinib as a possible candidate, something that likely would have taken months without the benefit of AI technology (the drug has been fast-tracked to clinical trials).
5. Reopening of economies
Although this one isn’t solely about healthcare, if there’s anything the past few months have taught us, it’s that the health of the population and the health of the economy are one and the same. With this in mind, governments are increasingly using AI applications to figure out how to balance public health and the economy, with the U.S. National Security Commission on Artificial Intelligence now calling for state governors to take a “data-driven approach” using health and other data when deciding how to reopen. And states are listening: In Pennsylvania, authorities are now taking advantage of risk-tiered metrics provided by multiple data sources to determine how quickly (and what) to reopen.
There’s no doubt the COVID-19 pandemic has been a catalyst for bringing AI and healthcare closer, and that this will only continue going forward. The only question now is just how much – and how quickly – AI and machine learning can help solve the mysteries of COVID-19, during a period when lives and economies hang in the balance and time is of the essence.