Asymptomatic carrier ratio estimation and asymptomtic transmission are becoming critical concerns in control of COVID-19 pandemic especially in consideration of re-opening.
Asymptomatic and Pre-Symptomatic Infection
Several studies have documented SARS-CoV-2 infection in patients who never develop symptoms (asymptomatic) and in patients not yet symptomatic (pre-symptomatic).1,2 Since asymptomatic persons are not routinely tested, the prevalence of asymptomatic infection and detection of pre-symptomatic infection is not well understood. One study found that as many as 13% of RT-PCR-confirmed cases of SARS-CoV-2 infection in children were asymptomatic.3 Another study of skilled nursing facility residents infected with SARS-CoV-2 from a healthcare worker demonstrated that half were asymptomatic or pre-symptomatic at the time of contact tracing evaluation and testing.4 Patients may have abnormalities on chest imaging before the onset of symptoms. Some data suggest that pre-symptomatic infection tended to be detected in younger individuals and was less likely to be associated with viral pneumonia.5,6
Asymptomatic and Pre-Symptomatic Transmission
Epidemiologic studies have documented SARS-CoV-2 transmission during the pre-symptomatic incubation period5, and asymptomatic transmission has been suggested in other reports.7,8 Virologic studies have also detected SARS-CoV-2 with RT-PCR low cycle thresholds, indicating larger quantities of viral RNA, and cultured viable virus among persons with asymptomatic and pre-symptomatic SARS-CoV-2 infection.9 The exact degree of SARS-CoV-2 viral RNA shedding that confers risk of transmission is not yet clear. Risk of transmission is thought to be greatest when patients are symptomatic since viral shedding is greatest at the time of symptom onset and declines over the course of several days to weeks.10 However, the proportion of SARS-CoV-2 transmission in the population due to asymptomatic or pre-symptomatic infection compared to symptomatic infection is unclear.11
- Chan JF, Yuan S, Kok KH, et al. A familial cluster of pneumonia associated with the 2019 novel coronavirus indicating person-to-person transmission: a study of a family cluster. Lancet 2020;395:514-23.
- Lu X, Zhang L, Du H, et al. SARS-CoV-2 Infection in Children. N Engl J Med 2020;382:1663-5.
- Dong Y, Mo X, Hu Y, et al. Epidemiology of COVID-19 Among Children in China. Pediatrics 2020.
- Kimball A, Hatfield KM, Arons M, et al. Asymptomatic and Presymptomatic SARS-CoV-2 Infections in Residents of a Long-Term Care Skilled Nursing Facility – King County, Washington, March 2020. MMWR Morb Mortal Wkly Rep 2020;69:377-81.
- Hu Z, Song C, Xu C, et al. Clinical characteristics of 24 asymptomatic infections with COVID-19 screened among close contacts in Nanjing, China. Sci China Life Sci 2020;63:706-11.
- Wang Y, Liu Y, Liu L, Wang X, Luo N, Ling L. Clinical outcome of 55 asymptomatic cases at the time of hospital admission infected with SARS-Coronavirus-2 in Shenzhen, China. J Infect Dis 2020.
- Pan X, Chen D, Xia Y, et al. Asymptomatic cases in a family cluster with SARS-CoV-2 infection. Lancet Infect Dis 2020;20:410-1.
- Bai Y, Yao L, Wei T, et al. Presumed Asymptomatic Carrier Transmission of COVID-19. JAMA 2020.
- Kam KQ, Yung CF, Cui L, et al. A Well Infant with Coronavirus Disease 2019 (COVID-19) with High Viral Load. Clin Infect Dis 2020.
- Zou L, Ruan F, Huang M, et al. SARS-CoV-2 Viral Load in Upper Respiratory Specimens of Infected Patients. N Engl J Med 2020;382:1177-9.
- Li R, Pei S, Chen B, et al. Substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (SARS-CoV-2). Science 2020;368:489-93.
July 10, 2020
Negative pressure rooms are designed to assure that the virus cannot escape the room. Chinese doctors analyzed hundreds of samples taken from various surfaces in such rooms in order to measure the amount of virus shed by the patients. Rooms of patients with symptoms ranging from negative to mild were found to be ‘extensively contaminated’ with COVID-19. However, the virus was not found in any of the air samples.
July 9, 2020
Many countries, including United States, struggle to contain the spread of coronavirus. Silent transmission of coronavirus is the transmission from the people who, at the time of transmission, are asymptomatic. The paper by Canadian and American scientists discusses the share of silent transmission in the spread of COVID-19 and the implications of silent transmission for the formulation of the most effective strategy to combat the pandemic. The results suggest that as many as 50% of infections originate from presymptomatic and asymptomatic people. Therefore, relying on traditional symptom-based interventions is not a plausible strategy. Extensive contact tracing and the widespread COVID-19 testing is required to stop the pandemic.
July 8, 2020
The paper published in Nature Medicine addresses the possible damage to lungs that occurs in patients that don’t ever experience symptoms and are not aware of being infected. These patients are truly asymptomatic, as opposed to the presymptomatic people who will eventually develop symptoms. The study shows that more than half of truly asymptomatic people have signs of inflammation in their lungs characteristic of COVID-19. Asymptomatic people do shed the virus; however to what extent, if any, they contribute to the spread of COVID-19 remains unclear.
July 7, 2020
Researchers from the Biomedical Research Center of Qatar University have analyzed 67 published studies of asymptomatic COVID-19 cases. While the percentage of asymptomatic patients vary from study to study, certain trends are apparent. Antibody tests show higher rates of asymptomatic patients than PCR-based tests. Young children rarely show any signs of the disease and remain asymptomatic, while continuing to shed the virus. Many adult patients who are initially asymptomatic eventually develop the symptoms. The most important question, how much the truly asymptomatic patients contribute to the spread of COVID-19, remains unanswered.
July 6, 2020
2000 individuals were randomly selected from the healthy population of Geneva and analyzed for the presence of anti-COVID-19 antibodies. The presence of antibodies indicated the presence of past COVID-19 infection. A five week study from April to May revealed the increase in positive tests from 5% to 11%. This is a much larger porportion of the population than the one based on the number of officially confirmed cases. On the other hand, the majority of the population remains unexposed to COVID-19, which means that the ‘herd immunity’ effect can’t stop the spread of the virus.
July 2, 2020
The article in News Medicine, published on June 18, addresses the important question of the meaning of the term ‘asymptomatic’ when applied to COVID-19 infection. It seems that ‘asymptomatic’ usually means the lack of complains by the patient who can continue hers of his normal life, despite the ongoing COVID-19 infection. They may be either unaware of any abnormalities in their body or consider the symptoms they experience trivial, for example tiredness or a headache. Studies show that under closer examination the majority of the asymptomatic patients have abnormalities in their lungs. About 80% of them have anti -SARS-CoV-2 antibodies. Many have some mild symptoms of infection.
July 1, 2020
Between March 11 and April, 9 there were no reported new cases of COVID-19 in Heilongjiang Province, China. Therefore when the new cases started to appear the authorities were able to trace their origins. It turned out that a single person, named A0 had originated a cluster of 71 sick people. A0 had returned from the USA, self-isolated for 14 days, showed no symptoms and tested negatively both for COVID-19 nucleic acids and antibodies. Despite all this, the downstairs neighbor who used the same elevator (not at the same time) was infected and started the chain reaction of infections. This is one of the rare cases when the infection from truly asymptomatic person was confirmed.
June 30, 2020
Doing more testing seems like a straightforward answer to pandemic, but including RT-PCR based tests are far from ideal: they produce large fractions of both false positive and false negative results. Currently, we don’t have a ‘gold standard’ to measure the quality of tests since a significant portion of people remains asymptomatic and there is no way to know for sure if they had ever been infected. Testing large pools of populations will lead to large number of false positive results. Mathematical modeling shows that with lower-end quality tests, as many as 20 uninfected people may have to be isolated in order to prevent one infection. This fraction sharply declines with the increase of the quality of the test. More accurate tests would provide an enormous health and economic benefit.
June 29, 2020
Presence of anti-COVID-19 antibodies in blood indicates that an individual is either currently infected or was infected in the past. Hameln-Pyrmont, the rural and sparsely populated region in Germany, reported an antibody-based COVID-19 rate of 0.889 per thousand, much less than that overall in Germany (1.99 per thousand). However, only 2.8 percent of the Hameln-Pyrmont population was tested for antibodies, which means that people who had no symptoms or mild symptoms were likely not counted. Doctors from Hannover Medical School set up the study of healthcare workers who were in frequent contact with COVID-19 patients in order to estimate the incidence of asymptomatic cases. They have found that 2.9 percent of staff have anti COVID-19 antibodies, indicating infection at some point of time, past or present. Only 36% of test-positive staff had experienced flu-like symptoms. The authors conclude that the real incidence of COVID-19 in the area is much higher than reported due to high incidence of people who are now aware of being infected. For that reason, the real mortality rate of COVID-19 is lower than reported.
June 26, 2020
Team of researchers from New York and Ohio was able to find the signs of COVID-19 infection in asymptomatic or presymtomatic volunteer students from small private college in Iowa. The researchers used the n-butanol sensitivity threshold test and so call Hedonics test (liking/disliking of various odors) for early detection of the changes in sense of smell. Participants who later developed COVID-19 symptoms had shown greatest changes in test results, even though they were not subjectively aware of any changes in their olfactory perception. Study demonstrates that olfactory test is superior to self-reporting of the same symptoms in early detection of COVID-19.
June 25, 2020
Long-term care facilities account for disproportionally large percentage of COVID-19 related deaths. Protecting the elderly living at these facilities requires vigilant monitoring of infections as well as better understanding of COVID-19 transmission, including asymptomatic transmission, specifically in cases when the elderly person is on the receiving end of transmission. Presymptomatic and asymptomatic workers are potential source of unrecognized COVID-19 transmissions. Scientists from Colorado State University and University of Colorado performed longitudinal surveillance of workers at skilled care facilities. The results of the study emphasize the importance of identifying and removing asymptomatic and presymptomatic workers from vulnerable populations.
June 24, 2020
Asymptomatic transmission is not uncommon among viruses. For example, the common flu virus can be transmitted from patients with no symptoms. Princeton scientists used mathematical modeling to demonstrate that asymptomatic transmission confers the evolutionary advantage to the virus, allowing it to spread ‘below the radar’. Symptomatic patients tend to self-isolate, for example, by staying at home while asymptomatic people continue to behave normally, coming into contact with more people. The researchers evolved their model by including patients who are mildly symptomatic and may or may not consider themselves sick. The researches had found that the most successful evolutionary strategy for a virus is to create a mix of asymptomatic, mildly symptomatic and symptomatic patients at early stages of the infection.
June 23, 2020
1573 healthcare workers, including physicians, nurses, healthcare technicians and clerical workers from University Hospital, Milan were monitored for the presence of SARS-CoV-2 infection. 139 of them eventually became infected. Out of 139 infected workers, only 17 had remained strictly asymptomatic, whereas the rest had developed some symptoms. Among the symptoms, the most indicative of COVID-19 were loss of taste and smell with fever being distant second.
June 22, 2020
Doctors from several Wuhan hospitals had monitored nine confirmed asymptomatic patients for the period of 85 days. During the study period none of the family members who lived in the close proximity of infected patients had become positive for COVID-19. Doctors propose that viral shedding may be less dangerous in the absence of sneezing and coughing. Also, patients and their cohabitants knew about the infection and presumably took steps to avoid the transmission.
June 19, 2020
Asymptomatic transmission remains a difficult challenge to address. Can all the cases of apparent asymptomatic transmission be attributed to transmission from the patients who develop COVID-19 symptoms a few days later? Is all ‘asymptomatic’ transmission really a presymptomatic transmission? The paper published in Nature Medicine takes a deeper look into the condition of asymptomatic patients. The study shows the presence of virus in nasopharyngeal swabs and urine samples of asymptomatic patients for many days, which means that these patients likely remain infectious.
June 18, 2020
The 3120 volunteers without self-reported COVID-19 symptoms from the Indianapolis metropolitan area were systematically tested by RT-PCR for the presence of COVID-19. About 3% of volunteers tested positive for COVID-19. Further detailed questionnaires revealed the presence of nonspecific symptoms in a fraction of the study group; for example, 10% had reported headaches. Other common nonspecific symptoms included muscle aches and shortness of breath. Overall, 72% of patients had remain asymptomatic while 28% had eventually developed some symptoms. The limitation of the study is that subjects were self-selected.
June 17., 2020
Asymptomatic transmission is perhaps the most popular topic of discussion around COVID-19. Dr. Natalie E. Dean explains why asymptomatic transmission is extremely difficult to detect in comparison to transmission detection from symptomatic COVID-19 patients. The very definition of ‘asymptomatic’ is unclear: people who appear to be fine can develop symptoms later, but there is no doubt presymptomatic people can transmit the disease. In regard to truly asymptomatic patients, Dr. Dean still advises caution: despite uncertainty, it is better to assume that asymptomatic transmission does occur.
June 15, 2020
The article presents the case of an asymptomatic patient who remained COVID-19 positive for 24 days under strict isolation conditions. Authors suggest that the period of COVID-19 related quarantine may need to be longer that 14 days.
June 12, 2020
Doctor Pablo M. Beldomenico suggests that propensity to become a COVID-19 superspreader may depend on the initial viral load: being infected with a large amount of the virus may cause the patient to produce more virus and become a superspreader. This domino effect may lead to an explosion of COVID-19 cases. While this is only a hypothesis, it may explain the dramatic differences in the rates of spreading of COVID-19 between countries and even within the same country at different times and locations.
June 11, 2020
Canadian clinicians discuss the steps that can be taken to decrease the risk of COVID-19 asymptomatic transmission to healthcare workers. Authors conclude that the risk of asymptomatic transmission is real. They outline the 5-level plan that includes switching to safer medical procedures, improving ventilation and using better personal protective equipment.
June 10, 2020
Doctor Chris Kenyon from Belgian Institute of Tropical Medicine argues that ‘super spreaders’, a small fraction of asymptomatic patients who shed high amounts of virus, play crucial role in COVID-19 infection. One prominent case identifies one person possibly causing 3900 infections. The study suggests re-thinking return to work strategies to emphasize the obligation to wear masks and avoiding gatherings in poorly ventilated areas.
June 9, 2020
Dr. Maria Van Kerkhove, the technical lead for coronavirus response at WHO, stated today that truly asymptomatic patients present relatively little risk of transmitting the disease. Many people who appear to be asymptomatic in fact do have mild symptoms or will develop a mild case of disease in a few days. Infected individuals may be able to transmit the disease two or tree days before showing any symptoms. Babak Javid, a principal investigator at Tsinghua University School of Medicine in Beijing, spoke to the distinction between asymptomatic and presymptomatc cases, “Other data available confirming that presymptomatic transmission does occur would suggest that being well does not necessarily mean one cannot transmit SARS-CoV-2.”
June 8, 2020
According to Prof Neil Hall, head of the Earlham Institute “Asymptomatic cases may be the ‘dark matter’ of the COVID-19 epidemic. As lockdown measures are eased, the importance of population-wide testing increases. Population needs to be made aware of dangers of asymptomatic transmission and the importance of following strict rules of social distancing.
June 5, 2020
Analysis of cases when COVID-19 had spread among isolated populations demonstrates that the incidence of asymptomatic people infected with SARS-CoV-2 is, by most conservative estimate, 30%. It was revealed that some people can feel no discomfort but still have noticeable damage to their lungs; others have no discernible damage. Both types seem to shed the virus and can infect others. The accumulated data suggests that testing only symptomatic people is a faulty strategy. Authors recommend increasing the number of tests by at least one order of magnitude compared to exiting level.
June 4, 2020
Scientists from Egypt have discovered a new symptom of COVID-19 that my be present in asymptomatic patients: a subtle loss of high frequency hearing. By measuring the audiological profiles of patients it’s possible to discover any damage to cochlear hair cells.
June 3, 2020
CDC had summarized available evidence on the existence of asymptomatic and pre-symptomatic transmission of COVID-19, including epidemiological evidence, virological evidence and evidence from modeling studies. The paper discusses public health implications of asymptomatic and pre-symptomatic transmission
June 2, 2020
Mounting body of evidence suggests that the main reason of why COVID-19 has spread with such speed, compared to other infections may be that people without the disease symptoms but infected by SARS-CoV-2 play a major role in spreading the disease. The paper suggests that symptom-based approach to testing must be replaced with mass testing, especially in nursing homes where concentration of vulnerable population is high.
June 1, 2020
As hospitals in Virginia begin re-opening for scheduled procedures many started testing all incoming patients, for example expecting mothers and people coming for elective surgery. The data sheds light on how many people in the general population are asymptomatic. Among pregnant women, the percentage of expecting mothers having COVID-19 was 1.6%; among people coming for surgery it was 1.8%.
May 29, 2020
Mounting evidence suggests that asymptomatic patients can transmit disease with high probability
Evaluating the degree at which asymptomatic patients can spread COVID-19 disease is complicated task; however there are known cases when infection was traced back to asymptomatic people. In some known cases one person who had no symptoms had infected several people. This presents unique challenge in stopping the COVID-19.
May 28, 2020
4 out of 5 people infected with SARS-CoV-2 on cruise ship were asymptomatic
The cruise ship departed from Argentina had 217 people on board. Out of 217 people 128 eventually got infected wish SARS-CoV-2; One person has died. 8 people (6.2%) required medical evacuation, 19,2% had symptoms. The majority, 81% were asymptomatic. For the first 28 days of the trip the ship, traveling near Antarctica, had no outside human contact.
May 25, 2020
The Time Scale of Asymptomatic Transmission Affects Estimates of Epidemic Potential in the COVID-19 Outbreak
This work described a mathematic model using generation interval of asymptomatic transmission to assess overall reproduction number.
May 22, 2020
Viral Dynamics in Asymptomatic Patients With COVID-19
This study found that in 31 patients virologically confirmed but were asymptomatic at time of admission, 22 presented symptoms afterwards. The rest 9 remained asymptomatic during hospitalization. Those 9 patients had low viral load but might still be capable of viral shedding and transmission.
May 19, 2020
Over 90% of Covid-19 cases in Michigan treatment center are asymptomatic
At two children’s treatment center in Michigan, 41 out of 44 residents who were tested positive for COVID-19 didn’t show symptoms. Another site none of 19 people with positive test results showed any symptom.
May 16, 2020
Cats With No Symptoms Spread Coronavirus to Other Cats, Lab Test Suggests
Researchers in University of Wisconsin School of Veterinary Medicine infected three cats with coronavirus from a human patient. Cats to Cats infection was also observed. But CDC said that based on the limited information, chance for pet to human transmission is considered to be low.
April 26, 2020
Most Americans Who Carry the Coronavirus Don’t Know It
An serological test indicated that as many as 2.7 million New Yorkers may have been infected without realizing it. This result was backed by several other observations in the world.
Artificial intelligence (AI) in the past 10 years gained a mainstream status as a technology. Today AI has become an important ally in the fight against coronavirus epidemic. There are several distinct areas where artificial intelligence has a potential to help in the fight against COVID-19.
Development of drugs and vaccines. Testing of new drug candidates is a tedious and expensive process that involves testing millions of chemical substances for their ability to inhibit the SARS-CoV-2 infection. Today, a significant part of searching for drugs can be done in silico, using computational modeling of the physical process of binding of the drug molecule to its target. AI is relatively new to this field, but significant progress has been made in this area. Based on examples of existing drugs that work against other diseases, AI can learn to recognize the chemical features of existing drugs that make the drug effective. Based on the ‘knowledge’ AI gains from the data describing existing drugs and the proteins they bind to, AI then can ‘look’ at new drug candidates and predict whether or not the candidate drug will work against COVID-19. To date, the accuracy of prediction has not been perfect, but using AI helps to prioritize the drug candidates for further testing in the laboratory.
Prediction of individual susceptibility to the virus. It is well known that many people infected with COVID-19 are asymptomatic, and showing immunity to the virus. Some have only mild symptoms, while others will develop severe – and sometimes fatal – symptoms. Overall about 80 percent of those infected have either mild symptoms or no symptoms at all. Presently we don’t have a clear picture to explain the vast difference in severity of infection. Being able to predict who is immune to the disease or would tolerate it with minimal symptoms would be enormously important. We could send some people right back to work or even recruit them to roles where they would be exposed to the virus. At the same time, we would be able to let the vulnerable part of the population stay home and wait until the epidemic is over or vaccines become available. Recently we have learned about the role of the ACE2 gene which encodes the protein responsible for the maintenance of the blood pressure. It appears that some mutations in the ACE2 gene, previously harmless and shared by many people worldwide, could facilitate the binding of SARS-CoV-2 to epithelial cells, making individuals carrying such mutations much more susceptible to the infection. There are most likely other genes that affect the susceptibility to the virus. Several large-scale studies are underway that aim to predict the susceptibility to SARS-CoV-2 by analyzing the individual genomes of the people.
Detection of infected individuals, contact tracing, hot spot identification. Artificial intelligence is capable of combining the information coming from different sources, such as self-reported symptoms, thermal imaging and other sources of data to estimate the probability that an individual has SARS-CoV-2 infection. Some countries, for example Israel, South Korea, Taiwan and United Kingdom utilize movement data from cell phones and video feed from street cameras or drones to track the spread of the disease. There is also substantial data coming from social media platforms. AI software is looking for posts about people experiencing symptoms, conversations about plans to get tested, or sharing the test results with friends and relatives. The AI can then analyze the stream of data and accurately predict emerging geographical hot spots of the disease. If used correctly the use of AI can alert the authorities of dangerous development and thus significantly decrease the number of infections and deaths by coronavirus.
Easing the load on healthcare professionals. The COVID-19 pandemic has created a worldwide shortage of healthcare personnel. AI can help to ease the load on doctors, nurses and other medial personnel by taking over some tasks that traditionally are performed by people. Remote monitoring devices continuously gather the information from the patients. AI processes the information and alerts the healthcare workers if the condition of the patient has changed. An additionalbenefit of remote monitoring is that it lessens the exposure of medial personnel to the infected patients who can spread the virus.
July 10, 2020
It is important to accurately differentiate the bacterial pneumonia from COVID-19 pneumonia. Engineers from Qassim University in Saudi Arabia used adversarial neural networks to teach the AI to make a distinction between the two. One of the networks, called ‘generator,’ creates fake chest X-Ray images based on bacterial pneumonia trying to fool the other network, called ‘discriminator,’ into classifying them as COVID-19 images. Discriminator’s goal is to make the correct diagnosis. Gradually, discriminator learns to pay attention to features characteristic of COVID-19 only, which improves the overall accuracy of the AI system.
July 9, 2020
AI based technologies are essential for extracting the real-time information about the coronavirus pandemic. At the same time, the widespread use of AI in combination with extensive surveillance measures raises privacy concerns. As an example, public transportation systems in China reportedly employ thermal scanning cameras in combination with face recognition to identify and track people with fever. The US CARES Act allocates $500 million to the development of public health surveillance and data collection systems. The Organization for Economic Co-operation and Development (OECD) had released the guidelines to help governments assure that privacy rights of citizens are not violated.
July 8, 2020
Researchers from Singapore created the Health Believe Model to obtain quantitative metrics of public attitudes toward the measures that Singapore authorities had imposed to curtail the spread of COVID-19, such as wearing masks and physical distancing. Deep learning-based text classifiers were able to classify the posts into four categories with accuracy over 90%. Analysis of Facebook posts revealed that relative frequency of comments in support of measures versus comments opposing measures change over time. The Singapore timeline showed two peaks when the number of relevant posts increased dramatically. The first peak corresponded to certain measures taken by Singapore authorities. The vast majority of posts were supportive of state-imposed policies. The second peak, a much larger one, had no clear origin and contained many posts complaining about restrictions. However, the overall attitudes of Singapore populations toward COVID-19 measures remain positive.
July 7, 2020
Amazon has unveiled the AI based monitoring tool that gives employees a real time warning about violating social distancing rules. Amazon likens this to radar speed check that gives drivers feedback on their driving. The Amazon tool, called Distance Assistant will guard against inadvertent or careless violations of six-feet distancing rules. Many had raised the concern that the employer gets the whole map of employee movements during the day, which then can be used for purposes unrelated to COVID-19 pandemic. Amazon plans to open-source this software.
July 6, 2020
Robots can replace humans in situations where the environment is dangerous to human life: on battlefields,nuclear disaster zones and terrorist threat situations. Increasingly, robots are used to perform tasks where there is high risk for exposure to COVID-19. Such tasks include the cleaning jobs in hospitals and places with high human traffic, the deliveries of food and medicines to sick people and the taking of vitals of COVID-19 patients. Simple task robots are relatively inexpensive: while the da Vinci surgical robot cost $3.5 million, the cost of disinfecting robots for hospitals is $30,000 to $80,000. In China, such simple-task robots are used to address the shortages of medical workers.
July 2, 2020
The US Department of State is responsible for carrying out US foreign policies and managing international relations. The COVID-19 pandemic created specific challenges for their function; one of them is the abundance of incomplete and outright false information about COVID-19. In response, the State Department is building a cloud-based AI platform called “Disinfo Cloud” to counter propaganda and misinformation.
July 1, 2020
Delays in detection of COVID-19 outbreaks are due to the combination of factors. One of the confounding factors is that the symptoms of COVID-19 overlap with other diseases such as influenza. A sophisticated AI system based on so called anomaly detection addresses this problem. While the symptoms may be similar on individual level, the frequencies of symptoms in population differ between COVID-19 and influenza. The AI is capable to detect the COVID-19 outbreak even it it overlaps with seasonal influenza outbreak.
June 30, 2020
There are currently 67,000 scientific papers about COVID-19. This this number is rapidly increasing and, as a result, getting answers to even simple questions is becoming a burdensome task. With this in mind, researchers from The Hong Kong University of Science and Technology in cooperation with World Health Organization’s Health Emergency Information and Risk Assessment Department have developed a publicly available AI-based question-answering system called CAiRE-COVID. In response to free from questions such as, ‘What are the risk factors for COVID-19?’ the system brings up relevant passages from multiple scientific papers. It also generates a brief summary to give users a concise answer to the question.
June 29, 2020
Door handles are contact points that are likely to be contaminated by the virus. Frequent cleaning of door handles helps to prevent the spread of COVID-19. To this end, engineers from Singapore University of Technology and Design and the University of Ontario have developed Human Support Robot with a customized cleaning module. The deep learning-based AI system helps the robot to recognize all kinds of door handles. After locating the handle, the robot uses liquid spray and a cleaning brush to clean up the handle. The tobot is equipped with the LiDAR (Light Detection and Ranging) system to navigate in the changing environment.
June 26, 2020
Unlike millions of other compounds discovered by chemists, FDA approved drugs had been already extensively tested for safety, which make them attractive candidates for fast-track development of anti COVID-19 drugs. Two AI algorithms were used to predict the binding of 2657 FDA approved drugs to ACE2, the protein that serves as a gateway for SARS-CoV-2 entry. Besides known ACE2 inhibitors the software had identified the drug involved in glucose homeostasis and the drug used against cystic fibrosis (a hereditary lung disfuntion) as anti COVID-19 candidates. Other candidates included drugs used for treatment of autoimmune diseases, obesity, hepatitis C and HIV.
June 25, 2020
Researchers from University of Michigan and Boston University measured protective effects of 1,425 drugs previously approved by FDA for treatment of various diseases. To measure the protective effect of drugs they used human cells grown in vitro as a model subject. They trained the Random Forest classifier to identify SARS-CoV-2 infected cells among millions of cells on tens of thousands of images. 660 unique features allowed AI software to identify and count infected cells on the images and thus to measure and compare the protective effect of drugs. Researchers had identified 132 drugs as having potential do protect human cells from being destroyed by SARS-CoV-2. The selected set includes some drugs previously indicated as promising by unrelated studies, for example remdesivir. These drugs are candidates for further investigation involving human subjects.
June 24, 2020
The drug combining antiviral and immunodepressant activities could be a powerful tool to fight COVID-19. Artificial Intelligence had predicted that baricitinib, an inhibitor of Janus kinase, can be effective against COVID-19. Since the drug was previously approved for rheumatoid arthritis, scientists had received the authorization to test it on four COVID-19 patients with bilateral pneumonia, one of them on ventilator. Within days, all four patients had improved, with both their viral loads and IL-6 levels (reflecting degree of inflammation) decreased. A much larger study, involving about 1000 patients, in underway.
June 23, 2020
Predictions made my AI are only as good as the available data. Data collection processes are not perfect and, as a result, minority groups can be either under-represented or over-represented in the data sets, which could lead to erroneous policy decisions. The article provides several examples of possible biases inherent in COVID-19 related data that can potentially lead policymakers in the wrong direction.
June 22, 2020
COVID-19 infection does not always result in visible changes in lungs, especially at early stages of infection. Despite that, CT scans remain to be a valuable diagnostic tool with a relatively small percentage of false positive results. It is fast and available in the majority of hospitals. The paper published in Nature Medicine shows that the accuracy of CT scan-based diagnostics can be improved by using Artificial Intelligence. In this study the performance of AI systems were compared to that of two trained radiologists. Two out of three AI system outperformed both radiologists in recognizing the COVID-19 infection on CT scans. The best performing AI model had combined CT scan images with clinical information.
June 19, 2020
There is an indisputable connection between big data analytics and the fight against pandemics, as pandemic response generates large amounts of data from the millions of patients. However, in practice the picture is much less clear. The first obstacle is the private ownership of the data. Medical institutions are interested in the results of big data analytics, but have little incentive to contribute their own data to common pools. There are also risks associated with potential breaches of the patient confidentiality. The second obstacle is the private ownership of the software. The creators of algorithms have little incentive to share the knowledge with the competitors. New approaches, both political and technological, are needed to overcome these obstacles.
June 18, 2020
The specifics of COVID-19 isolation policies differ between countries. Authors use three different AI algorithms to predict the future spread of COVID-19 in three countries (United States, Canada and Sweden) based on stringency of COVID-19 isolation policies. Models predict the possibility of a second wave of COVID-19 infection to impact the U.S. in New England and the Mountain states.
June 17, 2020
With more than 5000 COVID-19 publications weekly, keeping track of ongoing research is becoming an extremely difficult task. This is serious problem because new research is built on foundation of previous research. However, AI systems can be used to filter the literature and deliver more refined sets of publications tailored to fit the interests of individual researchers. SciSight, for example, is a free online tool that takes a sample paper and an input to display a browsable map or similar papers as an output.
June 16, 2020
The difficulties in data sharing such as privacy concerns and private ownership of the data is one of the biggest obstacles in applying the machine learning to biomedical domain. Federated learning, is the family of algorithms which can use decentralized data sources; the training occurs on separate data set and only intermediate results of training, but not the data is shared between participants. Several teams of scientists from China, Thailand and USA have build the AI application that recognizes COVID-19 on series of chest CT scans from different hospitals. Images themselves were not shared; instead the sets weights of the neural networks were averaged in synchronized learning cycles. The federated network performance was similar to the control performance on aggregated data
June 15, 2020
In order to develop immunity to the virus, the chunks of viral proteins must be first ‘presented’ to the immune system by special types of blood cells carrying so-called MHC receptors of their surface. Different chunks of viral proteins (called peptides) have a very different propensity to be captured and presented by MHC; therefore, predicting which peptides have a chance to induce strong immunity is important for developing a vaccine. Peptides causing the robust immune response can be mass-produced by the bio-pharma industry and made into efficient vaccines. An AI tool developed by scientists at MIT and Duke University had analyzed 150,000 peptides and selected 19 peptides to be synthesized in the laboratory for developing a vaccine against COVID-19.
June 12, 2020
Clinical data had confirmed that Baricitinib, a drug previously approved for treatment of rheumatoid arthritis, is effective against COVID-19. Earlier, this drug was proposed as a candidate for COVID-19 treatment by the researchers at Benevolent AI, a London startup specializing in AI.
June 11, 2020
A team of AI scientists from several US universities used the popular YOLO convolutional neural network to separate CT scans of COVID-19 patients from CT scans of patients with unrelated pulmonary diseases. The model had an accuracy of 97% and a sensitivity of 98%. The YOLO algorithm is able to highlight the features on CT scans that it had found to be characteristic of COVID-19 infection, which is important for understanding the pathology of COVID-19.
June 10, 2020
Taxonomic classification and elucidation of the origin of the virus is relevant to fight against COVID-19 pandemic. The researchers from the University of Western Ontario used novel, alignment free approach to produce highly accurate classification of 5000 viral genomes, including 29 SARS-CoV-2 genomes. The method, based on unsupervised machine learning helps to discover relationships of SARS-CoV-2 to other virus species.
June, 9, 2020
3D structures of some of SARS-CoV-2 proteins are not available from experiments, making it difficult to predict these structures using traditional 3D prediction models based on homology. Researchers from Michigan State University used machine learning tools trRosetta and AlphaFold in combination with molecular dynamics simulations to predict 3D structures of SARS-CoV-2 proteins.
June 5, 2020
Q-Learning, the algorithm that was used to build AI capable of playing computer games, turned out to be useful for drug design. Researches at Pucho Technology Information in Bangalore, India, have developed an AI system that can learn to ‘build’ the new drug candidates from groups of atoms much the same way as AI learns the right sequence moves in the game by trial and error. Combining AI with traditional physical modeling of drugs seems to be promising in creating new drugs against COVID-19.
June 4, 2020
A team of scientists at New York University created a mobile app that uses a combination of a patient’s medical history and blood tests to predict the severity of COVID-19 infection. The levels of certain factors in blood, such as C-reactive protein, myoglobin, procalcitonin and cardiac troponin, as well as data from electronic health records calculate a risk score ranging from 0 to 100. The instrument for quick analysis of blood factors – similar to an insulin measurement device – is being developed.
June 3, 2020
A team from Northwestern university has built an AI system that they believe can be trained to recognize promising high-quality research on COVID-19. The suggest it may be used in addition to human expert panels to allocate the societal resources for the fight against pandemic.
June 2, 2020
Knowledge graphs have become a popular way to capture the knowledge about the issue in a way that it can be explored by artificial intelligence. Benevolent AI, a London startup, has built the COVID-19 knowledge graph that encompasses much of what we know today about the SARS-CoV-2 virus. Based on this body of knowledge they were able to suggest that certain drug, known to medical professionals as janus kinase inhibitor baricitinib can be beneficial for treatment of COVID-19.
June 1, 2020
Data-driven predictions about the future course of COVID-19 epidemic are far from perfect, Errors in perdition may be due to unexpected events such as strengthening of restrictions in Singapore or protests in USA. Data science is, at it’s best, only as good as the data available; however, data driven predictions are better than any alternative.
May 29, 2020
Diveplane, an AI startup aims to solve the problem of tracing the people movement and spreading of COVID-19 without violating their privacy. At the core of their technology is ‘masking’ peoples identities by creating the ‘parallel database’ that contains all contact information but hides peoples identities, This database can be explored to study the spread of COVID-19 while protecting civil liberties.
May 28, 2020
AI gave an early warning on the cluster of unknown disease in December 2019
ProMED is a global disease monitoring system winch captures information on unexplained cases of the disease. On December 30, 2019 based on the analysis of social media it picked up a cluster of unusual pneumonia cases. Simultaneously, a disease surveillance system HealthMap collected the media alert about unknown respiratory illness. This illustrates that AI systems can be used successfully to identify and monitor the outbreaks of infectious diseases.
May 27, 2020
Chinese researchers built an AI model predicting the spread of COVID-19 based on the population flow from Wuhan province
Researchers team from China and Sweden had analysed over 14,000,000 data points related to the population movement from Wuhan to 296 prefectures in 31 provinces of China collected based on phone location records. They used machine learning to build the model that can accurately predict the spread of the virus based on the movement of population and the quarantine measures imposed by the state.
May 26, 2020
Models predict that age-based targeted lockdown may work better
It is a known fact that age is a major factor in COVID-19 fatality. Older people at at much higher risk of severe infection and death. It would logically follow that it is possible that the targeted isolation policy may work better that uniform isolation policy. However, it is not clear of exactly how such targeted isolation policy should be applied across different age group. It is very difficult to come up with clear cut guidelines for such policy, not to mention that it is not clear whether targeted isolation will work better than uniform isolation. In order to evaluate possible advantages of targeted isolation the scientists at National Bureau or Economic Research build a computational model that aims to predict infection rates, hospitalization and fatality rates under different lockdown scenarios – both uniform and age-targeted. They had shown that under some conditions the targeted lockdown works better than uniform lockdown. Of course the larger issue is how to enforce the targeted lockdown, given a fact that authorities have hard time enforcing even a uniform lockdown.
May 25, 2020
Artificial Intelligence-Enabled Rapid Diagnosis of Patients With COVID-19
This worked used AI to diagnose COVID-19 using CT scan and other clinical data. The method’s CT performance was as good as experienced senior thoracic radiologist and it was able to identify patients that had been tested negative using nucleic acid kit alone.
May 22, 2020
Researchers use AI to re-purpose drugs against COVID-19
Many drugs are found to be useful against more than one disease. This observation has led to the drug development strategy called repurposing. Existing drugs have been already tested for safety and lack of severe side-effects. This makes the path to finding drugs for new disease much shorter. Therefore often it makes sense to begin with testing the set of existing drugs before venturing into the unknown territory of new substances. The group of scientist from the Institute of Biotechnology and Pharmaceutical Research, Taiwan, trained neural network to recognize drugs that work on feline coronavirus based on data that had been collected in the past. The trained network was then used to predict the drugs that will work on SARS-CoV-2, which is similar to feline coronavirus. Neural network had identified 8 candidate drugs that are currently used against cancer, tuberculosis and other diseases as potential candidates for COVID-19 treatment.
May 21, 2020
Covid Scholar helps to search COVID-19 papers
|Scientists spend 23 percent of their time reading papers or searching for papers. Ever increasing volume of information requires new advanced tools that help scientists to find the relevant papers. Group of AI researchers at Lawrence Berkeley National Laboratory developed Covid Scholar, an online tool to explore scientific papers related to COVID-19. Instead of relying on papers containing same words AI looks for deeper connections between publications, for example when papers mention drugs with similar chemical structure. So far AI researchers amassed the collection of 60,000 papers, growing every day. The tool allows the lab scientists to find hidden gems that were missed by traditional PubMed searches.|
May 20, 2020
AI used to accelerate the design of T-cell vaccine
|Inducing T-cell immune response is very desirable for efficient vaccine. While T-cell immunity does not prevent virus entry into the cell it helps the body to recognize and destroy infected cells before virus has a chance to reproduce. T-cell response involves so call ‘presentation’ where parts of virus proteome in form of short peptides are presented on the surface of the T-cell. This is the mechanism by which the body can recognize the alien proteins and develop immune response to them. The peptides comprising virus proteome are not equal in their ability to be presented and to induce immune response. Some are poorly bind to the MHC (‘handles’ that presents them), others, when attached, do not induce sufficient immune response. Researchers from Norway and Germany used Artificial Intelligence to analyze the SARS-CoV-2 proteome and to calculate so called Antigen Presentation Score (APS) for each of the peptides. The peptides with highest APS were selected as candidates for the vaccine development.|
May 19, 2020
Urban monitoring predicts infection spread hot spots
Combining Artificial Intelligence with various sources of information ranging from aerial photography to smart phone ping data, commercial transactions data and social networks allows to measure peoples movements patterns throughout the city. This, in turn, allows to predict the infection hot spots where virus transmission is most likely to occur.
May 18, 2020
AI helps with remote monitoring of COVID-19 patients
A team led by MIT professor Dina Katabi has developed the AI-based system called Emerald. AI system analyses electromagnetic signals generated by people’s motion and is able to analyse breathing patterns. System is able to monitor patients located in separate hospital room of at remote location which greatly reduces risk to medical personnel and reduces the workload on doctors.