Advances in technology, such as artificial intelligence, could help to provide solutions to the COVID-19 pandemic. Using AI tools such as machine learning and data analytics, researchers can help to track outbreaks, detect illness and eventually secure a treatment.
However, putting in place the systems to do so will require time and may be more useful should another pandemic strike.
Tracking the outbreak
The Canadian AI company BlueDot reportedly noticed an unusual number of pneumonia cases in Wuhan, China. The company, which tracks outbreaks of infectious diseases, alerted its clients to the activity nine days before the World Health Organization officially flagged COVID-19. Using data analytics, BlueDot seeks to track, contextualize, and anticipate infectious disease risks.
Similarly, other companies such as HealthMap at Boston Children’s Hospital and Metabiota in San Fransisco also reported finding early signs of the outbreak. Metabiota has also proved adept at predicting the infection rate of the virus. In late February, it estimated that 127,000 people around the world would be infected by early March; this was only 30,000 more than the actual figure and remains within the margin of error.
However, as the pandemic spreads, reliable data will be more difficult to access thus making the predictions less accurate. News sources and official reports may offer inconsistent accounts, confusion will likely spread and data noise will be on the rise.
Detecting the illness
Some have claimed that AI could be used to provide early diagnosis of COVID-19 in patients. Artificial intelligence has been used successfully to examine medical images and catch early signs of breast cancer, and some researchers have suggested that machine learning could be used to detect COVID-19 in CT scans of lung tissue.
However, it is not clear whether sufficient data exists to train the AI models. In addition, the signs of the virus in the lungs may only be noticeable long after the initial infection and thus not applicable for an early diagnosis.
Another company, Infervision, has developed software which detects visual signs of the pneumonia. It has been used in several hospitals in China to help hospital staff prioritize patients.
Data is essential to find a treatment for the COVID-19 virus. For example, AI can be used to sift through existing drugs to see if they can be repurposed for the current virus.
AI could also be used to determine all of the possible evolution paths of the virus. Adding information about a possible vaccine could also help researchers better understand what would – or would not – work.
Research is also underway to understand the structure of the virus in order to create a vaccine. Because COVID-19 is a type of coronavirus that contains a single-strand RNA structure which allows it to mutate rapidly, the Chinese company Baidu has shared its Linearfold algorithm which can be used to predict the virus’s secondary RNA structure.
Such work, however, will require significant time. It is not expected that a vaccine will be developed for another 18 months.
Work in standardization
The IEC and ISO are developing international standards on AI through a joint committee (ISO/IEC JTC 1/SC 42) which is considering the entire AI ecosystem.
The SC 42 portfolio of deliverables covers foundational aspects such as a framework for AI using machine learning, standards lifecycle and terminology that enables the diverse stakeholders to engage and interact. These are in addition to horizontal deliverables such as data aspects, trustworthiness, computational methods, governance implications of AI, use cases and applications of AI.