When it comes to fighting the coronavirus pandemic, incorporating machine learning in public health might be one of our greatest tools. Machine learning and other artificial intelligence technologies are currently being developed and deployed worldwide to track, diagnose, and treat individuals with COVID-19.
Machine learning may not be the first solution that comes to mind for issues dealing with infectious disease. But this type of artificial intelligence, which can learn from past experiences and information, is incredibly well-poised to rapidly digest and analyze large amounts of data.
As the data on COVID-19 changes by the hour (or even minute), machine learning tools can process and integrate this data into their existing algorithms far faster than humans. Keep reading for a closer look at how organizations across the globe are using artificial intelligence in public health to fight against the coronavirus.
How Machine Learning and Public Health Are Tracking COVID-19
For scientists, journalists, and the public alike, one of the biggest challenges in understanding this pandemic and its impact has been accurately tracking the spread of COVID-19. Data on the number of cases in any given population, where it is available at all, changes by the minute. Researchers have been avidly searching for better ways to track COVID-19 cases and predict how the virus will continue to spread across the globe.
Predicting Individuals Vulnerable to COVID-19
Jvion, a company known for improving patient care through the use of calculated AI predictions, has produced a COVID Community Vulnerability Map. Using machine learning technology, Jvion is helping to determine which people and communities might be most vulnerable to severe outcomes, such as hospitalization or death, should they contract COVID-19.
In developing their algorithms, Jvion included not only traditional measurements of susceptibility to severe outcomes from COVID-19, such as chronic illnesses, but also typically overlooked social risk factors. These social risk factors range from lengthy commutes to living in highly populated residential areas, such as dorms or apartments.
Local government agencies, healthcare providers, and other decision makers can use this data to appropriately implement proactive approaches and allocate resources where they are needed most.
Predicting How COVID-19 Spreads Through Communities
Companies like Jvion are using AI to predict which individuals and communities would be the most vulnerable if COVID-19 were to spread in their region. Other companies are using AI to track existing cases and even predict where exactly COVID-19 will spread before it does so. At Entrust Solutions, we are proud to be developing machine learning solutions for precisely this purpose.
The software we are developing uses deep learning, a subset of machine learning, designed to both track where COVID-19 currently is and accurately predict where exactly the virus will spread. There are numerous similar programs already in existence, but what sets ours apart is our use of community structures in tracking and predicting the virus’ spread.
Led by Gene Locklear, Entrust’s Senior AI Research Scientist who is spearheading our innovative R&D efforts, our team is using a series of mathematical models based on studies of overlapping communities in complex networks. This involves teaching our software a nuanced understanding of how individuals interact with one another across multiple social networks. We aim to accomplish this by teaching the software several machine learning algorithms, such as hierarchical clustering.
To create this program, we plan to incorporate data mined from a variety of sources, ranging from social media platforms like Twitter to official CDC reports. We will also select data from previous epidemics as relevant, given that the body data on COVID-19 is still relatively small.
Our objective for this fusion of machine learning and public health is to ultimately provide both real-world analysis of COVID-19 as well as simulations of the virus spread vectors. We hope that this software will arm researchers, healthcare providers, and the general public with more data and knowledge to be as prepared as they can be.
How Machine Learning and Public Health Are Diagnosing COVID-19
Although numerous COVID-19 tests have been given emergency approval by the FDA, testing shortages remain a problem in the U.S. as well as worldwide. The reasons for these testing shortages are vast and complex, but include an exhausted global supply chain, the coordination among individuals and machinery involved in the current multi-step testing process, and the disconcerting rate of inaccurate test results.
Accurate and widespread testing is needed in order to know when it is safe to reopen our cities and countries. Even after reopening, the need for testing availability will continue until a vaccine becomes widely accessible. Testing also provides key information for tracking the virus and enables individuals to make informed decisions about their health.
That’s why an increasing number of researchers are combining machine learning and public health in an attempt to create COVID-19 diagnostics that are accurate, quick, and have the ability to be created and used in mass quantities.
Artificially Intelligent Cameras
Security cameras aren’t just useful for workplaces, government agencies, and airports looking to bulk up their defenses. Using machine learning capabilities, some security cameras can also help diagnose COVID-19.
None of these cameras function as substitutes for actual diagnostic tests. What these cameras can do, however, is rapidly scan individual temperatures using thermal technology. If the camera detects that someone has a higher temperature than normal, it could indicate that they have COVID-19.
Such cameras use facial recognition software developed through machine learning. By training the AI to recognize and distinguish between individual faces, these cameras can actually pick out which specific person within a crowd has a high temperature. This offers us far more valuable data than simply recording the presence of a high temperature somewhere in the vicinity.
Some of these cameras, such as those produced by the China-based company Hawang Technology Ltd., are attached to a separate temperature sensor. The combined temperature sensor and AI camera can then identify the person’s face and attach it to their name. The facial recognition software of this system is so accurate that it can even recognize people wearing face masks or coverings.
Other cameras, like the ones created by the German biometrics company DERMALOG, combine the temperature sensor and the camera into one product. These Fever Detection Cameras are actually able to read an individual’s temperature based on its facial recognition capabilities.
Both types of products can be used both to immediately detect the presence of someone with a fever, even within a crowd. These machine learning cameras could also be attached to automatic alert systems, which could notify the people identified with high temperatures.
Machine Learning and Image-Based Diagnostics
Another recent promising use of machine learning in public health is in the field of image-based diagnostics. Many researchers are combining machine learning and image-based diagnostics, such as CT and X-ray scans, to detect COVID-19 infections that could be more severe or life-threatening.
Numerous artificial intelligence initiatives are already underway:
- At Zhongnan Hospital in Wuhan, China, the radiology department uses AI to rapidly scan and analyze CT lung scans for signs of pneumonia that often indicate severe COVID-19.
- Remin Hospital, also located in Wuhan, has been examining CT scans with deep learning to discover COVID-19. Their technology has an accuracy rate of 95%, but has so far been used only to confirm diagnoses rather than detect new infections.
- An AI called COVID-Net has been created to detect COVID-19 by scanning chest X-rays. It uses data from X-rays of patients with a variety of lung conditions.
Although this technology has not yet been put into practice on a wide scale, researchers are excited by the possibility of this type of COVID-19 testing. Current COVID-19 tests are difficult for many hospitals to acquire and deploy as rapidly as they need.
Most hospitals, however, already have the machinery required for X-rays and CT scans on hand. If machine learning in image-based diagnostics proves to be a valuable resource for COVID-19 testing, hospitals could potentially code their own simple deep learning technology or leverage AI smartphone apps to detect COVID-19 in their patients.
How Machine Learning and Public Health Are Treating COVID-19
The particular strain of coronavirus we are facing was only identified a few months ago. This makes COVID-19 a relatively new virus in terms of existing data. Furthermore, the data that we do have occurs, and often changes, in real time.
The novelty of COVID-19 poses challenges to scientists and medical providers hoping to understand and treat this virus. However, by using machine learning in infectious disease research for COVID-19, we can greatly speed up the process of understanding COVID-19’s structure and possible treatments.
One of the pivotal steps in creating a vaccine is building an accurate model of COVID-19’s genome sequence. With the help of AI technology, scientists based in China were able to replicate the genome sequence of COVID-19 within a mere month. By comparison, it took several months for researchers to figure out the precise genome sequence of the 2003 SARS virus.
Meanwhile, scientists in Australia have been able to successfully create a lab-grown copy of the virus. This makes it far easier to test possible treatments in laboratory settings.
In addition, the AI company DeepMind has published a deep learning library named AlphaFold, which predicts the protein structures of various organisms, including COVID-19. The shape of an organism’s receptors is determined by the organism’s protein structures. Knowing how the COVID-19 receptors are shaped, or at least having a good working model, can help scientists develop drugs capable of bonding to the virus and disrupting its mechanisms.
Armed with knowledge about the structure of COVID-19, researchers worldwide have set out to discover possible treatments and vaccines for coronavirus. Broadly speaking, there are two approaches to treating COVID-19: finding an existing drug that fights against the virus or creating a new one.
Multiple companies and researchers have been utilizing machine learning in public health to discover existing drugs that could help fight COVID-19.
For instance, scientists at Deargen, a South Korean company dedicated to fighting diseases with AI, created a deep learning model that predicts how well a molecule will bind to a protein receptor. They found that a drug currently used to treat HIV shows great promise for treating COVID-19, too. Meanwhile, researchers at the Singapore-based firm Gero used deep neural networks to discover that a drug currently used for lung cancer treatment might offer relief from COVID-19.
One AI-discovered potential treatment for COVID-19 has already entered its clinical trial phase. The discovery that baricitinib, a drug currently FDA-approved for rheumatoid arthritis, could also help COVID-19 was made by a company based in England called BenevolentAI.
Although it had previously used its AI-powered knowledge graph to try to find treatments for chronic illnesses, BenevolentAI turned its attention to COVID-19 several months ago. BenevolentAI’s graph works by synthesizing large amounts of data from scientific papers and biomedical research. It then analyzes this data with the goal of finding connections between disease properties and drug structures.
This deep learning technique allowed BenevolentAI to identify a group of genes responsible for the parts of a cell through which the coronavirus can enter. As the company dug further into its research, BenevolentAI found results suggesting the anti-inflammatory properties of baricitinib could also help mitigate the respiratory problems in more severe COVID-19 patients.
While some scientists are working to discover existing drugs that could help treat COVID-19, others are trying to develop a novel drug aimed solely at mitigating or eradicating COVID-19.
One such initiative is the result of a recent collaboration between Itkos and SRI International. Itkos, an AI company, is using deep learning models to create virtual maps for completely unique drug molecules tailored specifically to fight COVID-19. Meanwhile, SRI, a research center, is using its automated synthetic chemistry system to figure out the optimal way to bring those new molecules into tangible existence. This partnership means that brand new drug candidates for COVID-19 can be tested in a mere 1–2 weeks.
Insilico Medicine, a company based in Hong Kong and devoted to the combination of artificial intelligence and public health, has taken a similar approach. The company uses a machine learning discovery platform to create thousands of new molecules designed to bind to a specific coronavirus protein. Insilico Medicine has already begun testing some of its molecules in collaboration with a pharmaceutical company.
How Can Technology Help You?
Technology is moving quickly to keep up with this pandemic. But the impact of COVID-19 extends beyond just public health, affecting a great number of industries and aspects of our lives.
In these trying times, many companies have had to rapidly alter their existing business structures. As organizations continue to adapt to their new local and global circumstances, they may have to incorporate new technology into their operations, shift their delivery methods of goods or services, or improve cybersecurity defenses and protocols for remote workers.
If your organization needs help managing the new or altered technology aspects of your operations, please contact Entrust Solutions for a free consultation about how our tech experts can help you.