Does your mother tongue make you more prone to covid-19?

A recent study by researchers at Moscow’s RUDN University on why covid-19 spreads more among certain populations has everyone talking. Since late last year, when cases of the novel coronavirus were first detected in China, researchers have looked at the pandemic from different viewpoints. The researchers at RUDN University believe that the number of covid-19 cases in a country could be related to the use of aspirated consonants in the primary language of communication there.

Their paper is based on a 2003 study by Japanese researcher Sakae Inouye, who had correlated the spread of the original SARS virus with the language spoken by people who were infected. A total of 8,000 cases were recorded in 26 countries during the SARS outbreak. That study compared Japanese and English, basing its case on the fact that the latter has more aspirated consonants—and when they are pronounced, small droplets are released from the speaker’s respiratory passage. The US accounted for 70 of those SARS cases, Japan had none. This is despite the fact that the number of Japanese tourists in China during the outbreak was much higher than travellers from the US.

A matter of language

The researchers at RUDN experimented with the same concept to check if it would hold true for covid-19’s spread. Their study, published in August in the journal Medical Hypotheses, used official data from 26 countries that had more than 1,000 registered cases of infection as of 23 March. Languages from these countries were divided into two groups depending on the presence or absence of aspirated consonants. “According to the scientists, although the groups did not show statistically significant differences, the countries that predominantly spoke the languages of the first group (with more aspirated consonants) had more cases of COVID-19: 255 per 1 million residents, as opposed to 206 cases in the second group," a release on the study explains. Despite experiment limitations, the researchers say this hypothesis might prove useful for epidemiologists.

To understand how aspirated consonants work, take a look at a video on the YouTube channel English Language Club, an online community. It is simple. Try keeping your hands close to your mouth when you speak—it could be any language. Some sounds create a puff of air, others don’t. In the English language, the consonants “p", “t" and “k" are aspirated when they appear at the start of words. When you enunciate words like “phone", “Twitter" or “quick", the sounds are accompanied by exhalation.

What about aspirated consonants in Hindi, where many consonants exist in pairs? Take, for instance, the consonant “k" in Hindi, which has an aspirated version “kh". Same for “g", “gh" and “c", “ch".

Let’s talk covid-19

Vocal biomarkers, used in recent years to detect mental and physical health conditions, are being studied in the context of covid-19 too. Researchers at the Indian Institute of Science (IISc), Bengaluru, for instance, are working on a project to develop a similar tool for covid-19. They call it “Project Coswara". Sriram Ganapathy, assistant professor at the department of electrical engineering at IISC and head of the LEAP (Learning and Extraction of Acoustic Patterns) lab, which specializes in information extraction from speech, uses the example of how modern smartphone cameras automatically detect facial features to explain “Coswara".

Most camera setups in phones use Artificial Intelligence models that are trained to look for facial features. “In a very similar fashion, the diagnostic tool in Coswara will attempt to identify the features of the breathing sounds, cough sounds and speech sounds that behave differently for a patient versus a normal healthy individual," explains Ganapathy on the project website. “The advantage is that the model in the tool would be able to learn the behavioural patterns from thousands of subjects and, therefore, has the potential to identify the specific patterns that are symptomatic of the virus," he adds.

On the left are cough samples from an adult who is covid-positive, and on the right, samples from a healthy individual. These were analyzed using a time and frequency analysis. A healthy cough data has a vibratory nature, with high frequency content. On the other hand, cough samples from a covid-positive subject tend to have low frequency content and a harmonic structure.

This tool, which could be developed as a web or mobile application, would use these vocal imprints to analyse and then give a predictive number or probability score on whether or not a person could be infected with covid-19. There are, of course, multiple stages to this whole project, which aims to supplement the current available ways of covid-19 testing, says Ganapathy. In the first stage, the Coswara team will collect health data, including sound recordings, from participants. This data will then be pooled and placed in different buckets based on their health conditions. The next step is to look for voice patterns. “For example, if the cough sound from a covid-positive subject has a shrill- sounding nature, we can separate this out and say that this is an indicator of the disease. Once these behavioural patterns are identified, the third stage is to build this diagnostic tool and verify its validity on a larger population," adds Ganapathy, who says the team is currently between stages 1 and 2.

They are working with two hospitals—from Mumbai and Mangaluru. There are challenges, says Ganapathy. “One of the key issues, which the doctors pointed out too, is what will happen with the participants that are covid positive but are asymptomatic? This is one of the questions we don’t have the answer to.... At the moment, we are treating that as a separate pool of data because we don’t know if the cough and respiratory sounds of such patients will have a vocal signature that confirms the presence of the virus."