3 Ways AI Is Tackling COVID-19
Artificial Intelligence (AI) is one of the most promising and talked about new technologies today. This post aims to highlight how machine learning algorithms are being used to build AI systems that can help us overcome some of the challenges brought about by COVID-19. The three areas we will focus on here are:
- Personalised care
- Vaccine design and discovery
- Risk management
“Machine learning is a field of computer science that aims to teach computers how to learn and act without being explicitly programmed. More specifically, machine learning is an approach to data analysis that involves building and adapting models, which allow programs to “learn” through experience.” [1]
Personalised Care
One of the most mysterious aspects of COVID-19 is the different ways, and varying levels, it can affect those infected with the virus. Some are not even aware they’re infected whilst others, even those categorised as ‘low risk’, require intensive care.
To better understand the care that will be needed for each COVID-19 patient, researchers at the University of Cambridge are using machine learning to analyse COVID-19 patient information from Public Health England. The type of data analysed includes basic depersonalised patient information, lab results, hospitalisation details, risk factors and previous outcomes. The system, named Adjutorium, then assigns patients risk scores based on their likelihood of ICU admission or ventilator usage. These are then aggregated across the hospital to give a picture of future demand on resources.
These new insights can then be used by healthcare professionals to help them answer questions such as:
- Which patients are most likely to need ventilators within a week?
- How many free ICU beds is this hospital likely to have in a week?
- Which of these two patients will get the most benefit from going on a ventilator today?
Impressively, with only access to relatively small amounts of training data the AI system performed more accurately than existing and widely-used survival analysis techniques such as Cox regression or well-known indexes such as the Charlson comorbidity index.
Vaccine Design and Discovery
For society to transition out of lockdown and away from social distancing and PPE guidelines, people must develop immunity to COVID-19. It is estimated that around 60% of the population must become immune to the virus to achieve herd immunity [2]. However, the World Health Organisation (WHO) predicts as few as 2–3% of people have been infected and therefore possess the antibodies necessary to give them immunity [3]. Consequently, if governments adopt the herd immunity strategy of allowing the virus to run through the population without a vaccine, they risk overwhelming their health system which could lead to rapidly increasing death tolls.
Achieving immunity by producing vaccines for COVID-19 is widely seen as the safest route back to normality. The caveat is that vaccines normally take many years — sometimes decades — to develop. Experts believe we may be able to reduce that timescale to 12–18 months with a coordinated global effort that harnesses the latest artificial intelligence (AI) and machine learning technologies
Machine learning algorithms are being used by vaccine developers to spot patterns in large datasets that humans would often find very difficult, if not impossible, to do.
Dr. Fast and Dr. Chen, two leading Stanford-trained PhDs, give the following example: “…immunologists have identified nearly one million protein fragments that are presented on a cell’s surface and visible to T-cells. However, no human eyes would be able to tell you whether this is true of SYGFQPTNGVGYQPY, a fragment from the novel coronavirus. On the other hand, a machine learning model can learn to answer this question from those million other examples, building an understanding of what patterns among the letters representing amino acids lead to a high likelihood of presentation.” [4]
Machine learning is thus a powerful tool for speeding up the discovery and design of vaccines against COVID-19. However, its potential is limited mostly by the quantity of training data available. Dr. Fast and Dr. Chen concluded that if we could build datasets of a similar size and diversity to those used to train voice and face recognition models, then AI and machine learning could achieve comparable success in vaccine design. It could do so by increasing confidence in the efficacy of vaccines before they are trialled and thus rapidly accelerate a typically slow feedback loop.
Risk management
Until a vaccine is found however, social distancing, the donning of masks and the management of capacity levels will have to remain at the core of government strategies to keep the infection rate of COVID-19 low. This will enable businesses and communities to re-open with a reasonable amount of managed safety. However, without proper monitoring and measuring of adherence to these guidelines, transmission risk levels will likely increase and cities and nations could face a dreaded second wave of COVID-19 cases.
This is why Shapes AI have developed an artificial intelligence (AI) platform that, in a privacy-first GDPR compliant manner, analyses camera / video feeds to automatically assess the extent to which preventive measures, including social distancing and PPE wearing, are being observed and highlight any hotspots of potential danger. This enables businesses and public authorities to respond immediately to risks and make optimal data-driven decisions on resource allocation and future guidelines. What’s more, by utilising the platform, businesses can help restore customer confidence in visiting their physical locations by presenting a real time ‘COVID safety score’. This metric is a reflection of the occupancy level of the store or facility, combined with the extent to which there is compliance with social distancing and mask wearing.
Conclusion
AI should not be seen as a silver bullet for solving the crisis caused by the coronavirus. Rather it is an assistive tool to amplify and augment human capabilities and efforts in dealing with COVID-19. This is because AI currently excels mostly at narrowly constrained tasks for which good amounts of quality data are available. However, the three applications of AI in response to COVID-19 described above, highlight the fact that AI can nevertheless be put to very effective use in addressing a variety of the crucial challenges society is facing in this pandemic.
