Predict an epidemic before it hits
The Canadian BlueDot algorithm was faster than experts in recognizing the threat from the latest coronavirus. He briefed his clients on the threat days before the U.S. Centers for Disease Control and Prevention (CDC) and the World Health Organization (WHO) sent official notices to the world.
Kamran Khan (1), physician, infectious disease specialist, founder and CEO of the program Blue Dot, explained in a press interview how this early warning system uses artificial intelligence, including natural language processing and machine learning, to track even one hundred contagious diseases at the same time. About 100 articles in 65 languages are analyzed daily.
1. Kamran Khan and a map showing the spread of the Wuhan coronavirus.
This data signals companies when to notify their customers of the potential presence and spread of an infectious disease. Other data, such as information about travel itineraries and flights, can help provide additional information about the likelihood of an outbreak developing.
The idea behind the BlueDot model is as follows. get information as soon as possible healthcare workers in the hope that they can diagnose – and, if necessary, isolate – infected and potentially contagious people at an early stage of the threat. Khan explains that the algorithm does not use social media data because it is "too chaotic". However, “official information is not always up to date,” he told Recode. And reaction time is what matters to successfully prevent an outbreak.
Khan was working as an infectious disease specialist in Toronto in 2003 when it happened. SARS epidemic. He wanted to develop a new way to keep track of these types of diseases. After testing several predictive programs, he launched BlueDot in 2014 and raised $9,4 million in funding for his project. The company currently employs forty employees, doctors and programmerswho are developing an analytical tool to track diseases.
After collecting the data and their initial selection, they enter the game Analysts. after epidemiologists They test the findings for scientific validity and then report back to government, business, and healthcare professionals. customers.
Khan added that his system could also use a range of other data, such as information about a particular area's climate, temperature, and even information about local livestock, to predict whether someone infected with the disease could cause an outbreak. He points out that as early as 2016, Blue-Dot was able to predict a Zika virus outbreak in Florida six months before it was actually recorded in the area.
The company operates in a similar way and using similar technologies. Metabiotmonitoring of the SARS epidemic. Its experts at one time found that the greatest risk of the emergence of this virus in Thailand, South Korea, Japan and Taiwan, and they did this more than a week before the announcement of cases in these countries. Some of their conclusions were drawn from the analysis of passenger flight data.
Metabiota, like BlueDot, uses natural language processing to evaluate potential disease reports, but is also working to develop the same technology for social media information.
Mark Gallivan, Metabiota's scientific director of data, explained to the media that online platforms and forums can signal the risk of an outbreak. Staff experts also say they can estimate the risk of a disease causing social and political upheaval based on information such as disease symptoms, mortality and treatment availability.
In the age of the Internet, everyone expects a quick, reliable and possibly legible visual presentation of information about the progress of the coronavirus epidemic, for example in the form of an updated map.
2. Johns Hopkins University Coronavirus 2019-nCoV Dashboard.
The Center for Systems Science and Engineering at Johns Hopkins University has developed perhaps the most famous coronavirus dashboard in the world (2). It also provided the complete dataset for download as a Google sheet. The map shows new cases, confirmed deaths and recoveries. The data used for visualization comes from a variety of sources, including the WHO, CDC, China CDC, NHC, and DXY, a Chinese website that aggregates NHC reports and real-time local CCDC situation reports.
Diagnostics in hours, not days
The world first heard about a new disease that appeared in Wuhan, China. 31th of December 2019 A week later, Chinese scientists announced that they had identified the culprit. The following week, German specialists developed the first diagnostic test (3). It's fast, much faster than in the days of SARS or similar epidemics before and after.
As early as the beginning of the last decade, scientists looking for some kind of dangerous virus had to grow it in animal cells in Petri dishes. You must have created enough viruses to make isolate DNA and read the genetic code through a process known as sequencing. However, in recent years, this technique has developed tremendously.
Scientists don't even need to grow the virus in cells anymore. They can directly detect very small amounts of viral DNA in a patient's lungs or blood secretions. And it takes hours, not days.
Work is underway to develop even faster and more convenient virus detection tools. Singapore-based Veredus Laboratories is working on a portable kit to detect, VereChip (4) will go on sale from February 1 this year. Efficient and portable solutions will also make it faster to identify those infected for proper medical care when deploying medical teams in the field, especially when hospitals are overcrowded.
Recent technological advances have made it possible to collect and share diagnostic results in near real time. Platform example from Quidel Sofia I system PCR10 Film Array BioFire companies providing rapid diagnostic tests for respiratory pathogens are immediately available through wireless connection to databases in the cloud.
The genome of the 2019-nCoV coronavirus (COVID-19) has been completely sequenced by Chinese scientists less than a month after the discovery of the first case. Nearly twenty more have been completed since the first sequencing. In comparison, the SARS virus epidemic began in late 2002, and its complete genome was not available until April 2003.
Genome sequencing is critical to the development of diagnostics and vaccines against this disease.
Hospital Innovation
5. Medical robot from Providence Regional Medical Center in Everett.
Unfortunately, the new coronavirus also threatens doctors. According to CNN, prevent the spread of coronavirus inside and outside the hospital, staff at Providence Regional Medical Center in Everett, Washington, use Work (5), which measures vital signs in an isolated patient and acts as a video conferencing platform. The machine is more than just a communicator on wheels with a built-in screen, but it does not completely eliminate human labor.
Nurses still have to enter the room with the patient. They also control a robot that will not be exposed to infection, at least biologically, so devices of this type will increasingly be used in the treatment of infectious diseases.
Of course, the rooms can be insulated, but you also need to ventilate so that you can breathe. This requires new ventilation systemspreventing the spread of microbes.
The Finnish company Genano (6), which developed these types of techniques, received an express order for medical institutions in China. The company's official statement states that the company has extensive experience in providing equipment to prevent the spread of infectious diseases in sterile and isolated hospital rooms. In previous years, she carried out, among other things, deliveries to medical institutions in Saudi Arabia during the MERS virus epidemic. Finnish devices for safe ventilation have also been delivered to the famous temporary hospital for people infected with the 2019-nCoV coronavirus in Wuhan, already built in ten days.
6. Diagram of the Genano system in the insulator
The patented technology used in the purifiers "eliminates and kills all airborne microbes such as viruses and bacteria," according to Genano. Capable of capturing fine particles as small as 3 nanometers, air purifiers do not have a mechanical filter to maintain, and the air is filtered by a strong electric field.
Another technical curiosity that emerged during the coronavirus outbreak was thermal scanners, used, among other things, people with a fever are picked up at Indian airports.
Internet - hurt or help?
Despite the huge wave of criticism for replication and dissemination, spreading misinformation and panic, social media tools have also played a positive role since the outbreak in China.
As reported, for example, by the Chinese technology site TMT Post, a social platform for mini-videos. douyin, which is the Chinese equivalent of the world-famous TikTok (7), has launched a special segment to process information about the spread of the coronavirus. Under the hashtag #FightPneumonia, publishes not only information from users, but also expert reports and advice.
In addition to raising awareness and spreading important information, Douyin also aims to serve as a support tool for doctors and medical staff fighting the virus, as well as infected patients. Analyst Daniel Ahmad tweeted that the app has launched a "Jiayou video effect" (meaning encouragement) that users should use to send positive messages in support of doctors, healthcare professionals, and patients. This type of content is also published by famous people, celebrities and so-called influencers.
Today, it is believed that a careful study of health-related social media trends could greatly help scientists and public health authorities to better recognize and understand the mechanisms of disease transmission between people.
Partly because social media tends to be "highly contextual and increasingly hyperlocal," he told The Atlantic in 2016. Marseille Salad, a researcher at the Federal Polytechnic School in Lausanne, Switzerland, and an expert in a growing field that scientists call "Digital Epidemiology". However, for now, he added, researchers are still rather trying to understand whether social media is talking about health problems that actually reflect epidemiological phenomena or not (8).
8. The Chinese take selfies with masks on.
The results of the first experiments in this respect are unclear. Already in 2008, Google engineers launched a disease prediction tool - Google Flu Trends (GFT). The company planned to use it to analyze Google search engine data for symptoms and signal words. At the time, she hoped the results would be used to accurately and immediately recognize the “outlines” of influenza and dengue outbreaks — two weeks earlier than the U.S. Centers for Disease Control and Prevention. (CDC), whose research is considered the best standard in the field. However, Google's results on early Internet signal-based diagnosis of influenza in the US and later malaria in Thailand were deemed too inaccurate.
Techniques and systems that “predict” various events, incl. such as the explosion of riots or epidemics, Microsoft has also worked, which in 2013, together with the Israeli Technion Institute, launched a disaster prediction program based on the analysis of media content. With the help of vivisection of multilingual headlines, "computer intelligence" had to recognize social threats.
The scientists examined certain sequences of events, such as information about the drought in Angola, which gave rise to predictions in forecasting systems about a possible epidemic of cholera, as they found a connection between drought and an increase in the incidence of the disease. The framework of the system was created on the basis of the analysis of archival publications of the New York Times, starting in 1986. Further development and the process of machine learning involved the use of new Internet resources.
So far, based on the success of BlueDot and Metabiota in epidemiological forecasting, one may be tempted to conclude that an accurate prediction is possible primarily on the basis of "qualified" data, i.e. professional, verified, specialized sources, not the chaos of Internet and portal communities.
But maybe it's all about smarter algorithms and better machine learning?
