The Data Race Against The Pandemic, Together
How ML and data were crucial in fighting COVID-19 in 2020 as a united global community. 2020 was a year...
AI-based solutions are being used to detect outbreaks of infectious diseases around the world. To illustrate, HealthMap and BlueDot are two platforms that were early reporters of the unusually high number of pneumonia cases in Wuhan, at the end of December 2019. Such tools rely on advanced analytics and Natural Language Processing techniques to process real-time information spreading in the online world. By mining various sources of information such as online news sources, official medical reports, tweets, google queries, and even blogs and chat rooms, they can detect significant disease events.
Being able to label such events as signs of emerging epidemics is an effort for which both algorithms and human expertise are needed. The AI tools give medical experts a head start in the right direction. This is a powerful way to leverage real-time data and enable early prevention measures. And early prevention saves lives!
Let us continue with the example of the coronavirus epidemic. With confirmed cases first centered around a specific country, tools are needed to be able to track and monitor the spread of the disease across the world. But since humans are unpredictable beings that like to roam freely, forecasting where the virus will hit next is not an easy task. AI-based solutions (such as Metabiota) deal with air-travel data, real-time diagnosis information from across the world and hospital admission reports. Analyzing data coming from such a variety of sources is something that advanced predictive tools are designed for. This makes it possible to obtain accurate real-time insights (such as prediction of cities on the virus trajectory).
But again, expert knowledge is needed to augment the information provided by the AI systems. Having the capacity to exploit epidemic data and to extract insights is just the first step. Thanks to AI tools, health and government officials have accurate insights available for better decision making.
The real-time capacities of AI models gained them some attention in crisis response and management solutions. For example, the Qatar Computing Research Institute has an entire Crisis Computing team tasked with finding modern solutions to humanitarian emergencies. By combining Natural Language Processing and Computer Vision techniques, they analyze social media in order to develop situational awareness models. The integration of multiple analysis methods is needed in order to account for the variety of data sources and formats available. This provides relief organizations with 24/7 support and the opportunity to better target their response efforts.
Microsoft’s AI for Humanitarian Action program is another example of how technology and human expertise can come together to solve pressing issues around the world. When data modeling tools and AI algorithms are widely available, preventing devastating consequences becomes possible.
AI tools have a proven track record in medical use cases. There are two main directions in which AI superpowers have been historically leveraged against diseases: diagnosis and treatment.
Deep learning solutions are commonly used in medical imaging since they can identify radiographic changes in X-ray scans and CT images. When enough training data is available, they can identify specific radiographic features and correlate them with signs of disease. Such tools can assist medical staff on the time-consuming and error-prone task of patient screening.
For the purpose of developing treatment, AI tools have been successfully used for drug discovery. They are able to generate large numbers of molecular structures and search for the one best fitting for a potential vaccine. By leveraging existing molecular structures from similar viral diseases, the expensive process of creating vaccines could be drastically improved.
As of today, a number of research works are using AI tools for tackling diagnosis (here, here, here and here) and vaccine development (here) against the novel coronavirus. While most of the works still need to be peer-reviewed and prove their usability, such interest from the research community shows the great potential of using AI for tackling epidemics.
For more ways data science can optimize supply chain see our Case Study: