4 Ways Machine Learning Can Help Combat COVID-19
At this time, there is no bigger concern for healthcare systems, medical practitioners, and researchers than COVID-19. Most healthcare systems across the world have shifted their focus to combat the virus and are doing their best to ensure the health and safety of their patients.
Organizations have been leveraging various digital solutions to combat this outbreak, and one with high potential is machine learning (ML). ML uses large volumes of data to identify patterns to give decision-makers better insights to make better-informed decisions.
Picture Credit: Unsplash
As quoted in Health IT Analytics, “With the right data, integration methods, and personnel in place, machine learning has the potential to advance clinical decision support and help providers deliver optimal care.”
Here are four ways machine learning can be utilized to help fight COVID-19:
1. Machine Learning Helps Find Those That Are Most at Risk
Recently, new information became available about the specific effects of COVID-19 inside the human body. This gives data analysts a chance to create data clustering models that can showcase an interactive and simplified way to identify populations most vulnerable to the disease and their location. This allows healthcare personnel and supplies to be distributed to the areas most in need.
ML can also analyze the data in a growing hub of infections and estimate the infection risk for the infected person’s closest connections based on the interactions between them.
Another, more ambitious solution, could create a model that can help predict the result of a specific treatment in a group of people with specific symptoms, and accordingly, doctors can prescribe a treatment plan more efficiently with greater success because they would be able to predict the outcome.
2. Machine Learning Can Increase the Number of Ways COVID-19 is Pre-Diagnosed
The most common way to diagnose the virus is a PCR (Polymerase Chain Reaction) test, but because of the rapid increase in the number of cases all over the world, most countries do not have the proper infrastructure to support the demand of tests needed per day. According to a statement from WHO, “There is currently no evidence that people who have recovered from COVID-19 and have antibodies are protected from a second infection.” There needs to be a way to increase the number of tests available.
If a machine learning model could be created to use alternative factors to accurately pre-diagnose COVID-19, such as face scanned pictures or anomalies in a heart-rate evaluation, these alternative diagnoses could greatly reduce the need for physical interaction between two or more people and lower the risk of infection for health workers. Even if the accuracy is not high, this heuristic method can be effective in early diagnosis. One promising study used artificial intelligence (AI) algorithms to integrate chest CT findings with clinical symptoms, exposure history, and laboratory testing to rapidly diagnose patients who are positive for COVID-19. This could prove a valuable triage tool, especially while the PCR test supply is low.
Another possible solution is to develop an ML-enabled chatbot to help patients with standardized symptoms receive a contactless pre-diagnosis that also offers instant COVID-19 related information to enhance patient care.
3. Machine Learning Helps Understand the Nature of the Virus and How It Spreads
One of the most important things healthcare professionals are focused on is minimizing the impact of the pandemic by understanding how COVID-19 interacts with different bodies during infection; this is also known as Virus-Host Interactome. The main goal of mapping all these interactions is to develop new drugs or vaccines that can fight the virus.
In the past, effective drugs were developed against H1N1 using a map of interactions. Currently, there are many machine learning projects running all over the world to learn how the proteins inside our bodies interact with the virus. In one example, NVIDIA uses a graphics processing unit (GPU) deep learning to launch its Folding@home initiative. This initiative repurposes unused, community-sourced, GPU time to run machine learning simulations on COVID-19’s viral protein. Folding@home’s goal is to learn enough about the virus to design therapeutics to counteract it.
NVIDIA’s 2020 keynote revealed even more projects related to the use of machine learning in the battle against COVID-19. One of the most important ones was the first 3D recreation of the virus using all known interactions and proteins. The 3D recreation of the virus helps researchers learn more about the spike proteins of the virus and how it replicates into different shapes, which could be helpful for them to develop the potential vaccine and antibodies.
4. Machine Learning Helps Estimate the Projection and Prevention of Future Pandemics
Having a projection in the number of cases, deaths, and people at risk can be fundamental for any country in the world right now. Creating a prediction model using machine learning that takes all the data related to COVID-19 cases with additional information such as social behavior per city and social network interaction, can potentially give a view of what is to come in the near future. It is also a powerful tool for governments, and health leaders to make a timely decision to reduce the impact of the pandemic in society.
Finally, using the learnings from this experience, a new study of potential pandemics can be implemented. A list of common illnesses that affect specific types of animals already exists. In the past, some of those illnesses affected a human host and created a situation like this current one. A good approach to avoid it is to create machine learning models that can study the proteins and interactions of those potential illnesses to find a way to isolate the proteins that can migrate or interact negatively with humans.
Machine Learning Can Be a Vital Player in Combating COVID-19
In these difficult moments, everyone needs to work together to get through this situation and learn from it to prepare for future scenarios. Machine learning methods are one of the top research topics in the world, and hardware companies are constantly developing powerful devices that can run more complex algorithms to provide better insights. We should leverage these powerful technologies to help keep everyone safe and healthy.
How Nisum Can Help
Nisum has always been committed to serving our global and local communities, and right now, there is no greater crisis impacting the world than COVID-19. Now, we are actively joining the fight by offering our services free of charge to organizations that have shifted focus to actively fight against this pandemic. Visit our COVID-19 Pro Bono Offer page for more information.
About the Author: Humberto Rodrigues is a Full Stack Developer at Nisum. He is currently developing his M.Sc. in Computer Science related to the machine learning processes.