Streaming issues? Report here
Clement Manyathela 1500 x 1500 2020 Clement Manyathela 1500 x 1500 2020
The Clement Manyathela Show
09:00 - 12:00

Up Next: The Midday Report with Mandy Wiener
See full line-up
The Clement Manyathela Show
09:00 - 12:00
Business Unusual

Loss of smell is a common part of Covid-19, smell might help us fight it too

9 June 2021 7:15 PM

Dogs, bees and AI may help us find more infections faster than current testing

This introduction is very optimistic that some experimental results for using smell to detect disease will not only be faster and more effective than current methods, it will be a lot cheaper too.

Detecting for SARS-Cov-2 currently requires a very sophisticated test to determine if the viral replication code of the virus is present in a suspected infected person.

Polymerase Chain Reaction (PCR) is an incredible method to rapidly and relatively accurately duplicate trace amounts of DNA in a sample to determine if the sample contained the target virus.

South Africa has carried out almost 12 million tests (by June 2021) to detect the presence of the current strain of coronavirus and its many variants.

It is an incredibly useful test that was pioneered in the early 80’s and has allowed us to slow the spread of the virus more than would have been possible during the Spanish flu pandemic in 1918. Then, you would only know you have the disease from the side effects, those that were infected and contagious but not symptomatic would unknowingly spread the disease far and wide.

That may still be the case now, but at least the PCR test can confirm an actual infection and when done using random samples it can detect asymptomatic cases.

Tests typically have an accuracy rating to determine the likelihood it would correctly test for a specific condition and a corresponding likelihood that it would not detect the specific condition and a third factor which is the likelihood it would detect the specific condition incorrectly.

For PCR tests this is about 70% accurate with a corresponding 30% chance it will not detect an actual positive case and then a very time dependent chance of detecting a positive result that is incorrect from about 70% to 20%. The sooner the test is done the greater the chance it may incorrectly give a positive result that is false. The test needs to be done in a lab and takes hours to complete.

Antigen tests are far quicker and don’t need lab environments to do the test and cheaper. They are less accurate though, typically around 50%.

The point of this is that testing is not 100% accurate and so what you want is something that is accurate enough, fast and cheap and if needed can be referred for a slower and more costly test.

Only doing antigen tests for people arriving at airports would keep people moving but potentially allow new variants to enter the local population.

Doing random antigen tests offer a sense of what asymptomatic cases may be present with a percentage of the positive and negative antigen tests sent for more robust testing to check the accuracy for a particular location and its accuracy over time.

Ideally an instant test that could find many of the definite positive cases would ensure fewer get through with the more robust tests used for the occasions when the instant tests were inconclusive.

There is good reason to believe that the sensor that can work almost instantly, cost very little, does not need lab conditions and can be used repeatedly may be right under our noses.

We can smell and recognise over 10 000 unique smells, not that you have made a list. Yet we take it for granted even as it quickly and effectively allows us to find or avoid so many issues.

An even more effective solution can be found in the snouts and noses of fellow organisms that are far more sensitive.

Dogs are way better than us, they take only a little training to associate a smell with a reward and are about 70% accurate

Mice can smell better than us and as well as dogs, but are way smaller and easier to look after, rats are already a very effective way to sniff out landmines, a rat recently retired after having cleared 400 soccer pitches worth of minefields over 5 years

Bees have an very keen sense for detecting volatile chemicals and can be trained in minutes to reach an almost 98% accuracy using just breath for some conditions

A specific kind of wasp is even more adept at spotting a range of smells and even easier to train, the catch is that bees and wasps don’t live very long so fresh recruits are needed constantly.

The best sniffer by olfactory receptors has 1948 to our more modest 396 and is the African Elephant, maybe not too surprising given the giant nose if we can detect 10 000 smells often at a good range, then elephants can basically see with their telescopic trunks.

The elephant in the room for using them to detect disease is quite literally the elephant in the room.

The tests on Covid-19 are encouraging with the tests using dogs suggesting that besides sniffing drugs at airports they could take care of disease too.

Helsinki was the first to try and Singapore will be the next to test for this. Once again man's best friend proves just what a good friend he is.

But no sooner had dogs began to wag their tails in appreciation than a new contender arrived to challenge their status. It does not eat or sleep and lives longer than any dog - could we get a machine to sniff out Covid?

The challenge was to know what Covid smells like. Dogs and bees and all manner of other animals know, but we lack the means to ask them to describe it.

Rather than give up we simply turned to another very usual friend we don’t fully understand. Machine learning.

If we could be sure that we could create a smell sensor with enough sensitivity we need only expose it to the samples we know are Covid positive and then let machine learning get to work determining what it is that is present in all Covid cases but absent in those without it.

It is a crazy idea to ask a machine to effectively work it out for itself, but it appears to be showing good results. The machine using what it can detect has been able to match the sensing ability of dogs which can detect concentrations of a few parts per billion.

Consider your ability to taste whether your cup of coffee has any sugar in it. A dog could determine if there was a spoon of sugar when diluted into 5 million litres of water. That is the equivalent of two swimming pools of water. This is a dilution comparison rather than to conflate how we taste for something versus how we smell it.

It is worth noting that while we have maps of the paths and the cells that do the smell detection, we are still not sure just how we tell what we are smelling. The smell you recognise as freshly baked bread is a broad range of different compounds and our smelling cells only respond to some of them, yet somehow it gets detected and combined to trigger a smell reaction which in the case of smells we associate with danger gets us moving away from it before we even consciously are aware of.

If you have ever wondered why memory and emotion are often triggered by smell it is in part because it is wired to be processed by the same part of the brain.

In a previous episode that predicted future homes kitted with sophisticated chemical detection units that might be built into toilet to test our urine for a range of issues may be even more ubiquitous using a smell detector in the bathroom or even better building it into the mic slot on your phone or smart watch or glasses needing only a breath to check you are fit and healthy.

Imagine a smart speaker with the sensor, you need only ask it how you are doing and it would be able to tell in the same breath that you asked the question.

The challenge like many of these developments is not in discovering the theory, but building the units at scale with affordable prices and the kind of accuracy and robustness to last.

The progress can be slow and it may even seem to fail, but consider that in 1973 the global count of mobile phones was 1. Fast forward to 2012 and there were a billion, this year we will hit four billion.

This may be another case of what we are able to achieve from one year to the next might seem underwhelming, but what that amounts to in a decade is miraculous.

9 June 2021 7:15 PM

More from Business Unusual