AI School Blog
4 min read
It is no secret that COVID-19 has had a devastating impact on people’s lives. What you may not know, however, is that artificial intelligence has played a significant role in predicting the outbreak of the disease from the very beginning. Not only that, AI continues to shape public health responses worldwide. From fields such as diagnosis and treatment to drug development, AI technologies to combat COVID-19 have been versatile across many usage areas.
Weeks before public health authorities and global news outlets caught on to the severity of the problem, a Canadian digital health company built an algorithm based on machine learning and natural language processing to track the spread of the disease.Their algorithm compiled keywords from newspaper sources and flight itinerary data to warn their customers of the disease’s potential to turn into a pandemic. It also identified possible locations for the spread of the disease, accurately predicting 8 out of the ten first cities worldwide to have imported cases of COVID-19.
In order words, through the simple application of a machine learning model, the company could predict the outbreak of a major pandemic that was to have wide-ranging repercussions for the world as we knew it - before significant institutions such as the WHO started giving out a single warning sign.
Developing COVID-19 testing kits
In February of last year, South Korea was the country with the highest amount of recorded COVID-19 cases outside of China - although, as we all know, that trend certainly changed as more countries ramped up their testing just a few weeks later. Infection rates started booming in Europe, the US and Latin America.
Instead, South Korea is now renowned for its response to the virus with surprisingly low numbers of confirmed cases. This is remarkable if we consider their numbers of infected in proportion to their high testing rate - all without ever introducing a full-scale lockdown.
Wonder how they did it? They used AI-powered testing kits.
In the early days of the pandemic, South Korean biotech firm Seegene broke records when they used an AI computational model to develop test kits based on genetic data that had been released about the virus - all in just three weeks. Developing such test kits manually would have taken 2-3 months. Moreover, the use of automated testing methods by placing the samples into a diagnostic machine has also boosted the time it takes to receive the result by four times compared to the manual method.
More recently, the same company also deployed artificial intelligence to create the world’s first COVID-19 mutant test in another boost to the world’s testing capacity.
Antibody drugs are medicines that treat or prevent severe cases of COVID-19, use virus antibodies to neutralize areas of infection based on genetic engineering. As the same method has been successfully used to treat other viral diseases such as hepatitis C and HIV, the development of antibody drugs was quickly identified as a possible approach to combat COVID-19. The only problem was the time and capacity needed to manufacture such drugs, not to mention the capability to determine the best antibodies to use.
Artificial intelligence might have the solution, once again. Scientists from ETZ Zürich in Switzerland have recently developed a machine learning model using the DNA sequences of 40,000 antibodies. Their model was applied to a database of 70 million antibody DNA sequences to predict how well the antibodies would bind to the target protein to neutralize the virus.
The Zürich-based scientists narrowed the optimal DNA sequences down to 55 after applying further computer models, which they then used to produce antibodies. Later trials revealed that the AI-optimized DNA sequences performed better in binding to the target proteins than previous drugs - thereby highlighting the usefulness of AI-based methods in the development of antibody drugs to combat COVID-19.
Diagnosing suspected cases
Another popular use of artificial intelligence to combat COVID-19 is the detection and diagnosis of infections. Using machine learning classification models trained using clinical data and chest CT scans, medical institutions all over the world have been able to detect cases of COVID-19 with an overall accuracy exceeding 90 percent.
For example, a study conducted in early 2020 on 18 medical centers across 13 Chinese provinces revealed that an AI algorithm applied to the CT scans of 279 patients had an equal accuracy rate compared to that of a senior radiologist. Not only that, while radiologists incorrectly classified some patients who were positive for COVID-19 as negative, the AI system was able to correctly identify 68 % of these patients as positive. This example shows how AI as a tool can either be equal to or even outclass experts - and highlights the importance of AI in the healthcare sector and beyond.
For those interested, some of the datasets used to train machine learning models that can be used to classify the CT scans of COVID-19 patients are available publicly on Kaggle, Github , and MedicalSegmentation. If you are wondering how to train a machine learning model using AutoML, feel free to check out AIBrain’s online course, AutoML for AI-powered professionals. If you are interested in taking the next step to a career in AI, we have an early bird discount running until 6 June 2021. Maybe this blog could even provide some inspiration for your capstone project?
This blog article has identified popular usage areas of artificial intelligence to combat the pandemic. As artificial intelligence becomes more and more widespread in different fields, it is even more essential to build AI literacy worldwide. In doing so, we have to make sure that knowledge and the opportunity to build skills within AI are accessible to those who desire it.
While debates are running hot over the potential problems of AI in increasing automation and reducing jobs, humans are the source of creative intelligence at the end of the day. That is why AIBrain’s vision is to augment human intelligence with AI - and we aim to work harder every day to make sure that AI is accessible to everyone.