Ann: US COVID-19 study completes recruitment, page-14

This study was to collect training data. Now they have to find an acoustic signature (or combination of signatures) that are predictive of Covid-19 when used with a particular machine learning algorithm. This can be quite time-consuming as there are many signatures that can be extracted from an acoustic recording, and there are also many machine learning algorithms. They will also have to use their existing training data to make sure the signature(s) and algorithm exclude coughs from the other target conditions.

Once the right combination has been discovered it must be tested in a predictive trial. If the company is prepared to spend the money on a full-blown arms-length, double-blind trial then if the results are sufficiently good then the results can be used to support regulatory approval.

There is good reason to be very cautious here:

• Is it possible to distinguish between Covid-19 lung alteration and some of the other conditions that alter lung tissue? But at least the company possesses training data that will allow them to test this question.
• What was the dominant variant in the Covid-19 cough sounds they collected? Hopefully the company captured this information.
• Does the omicron variant produce the same type and degree of lung tissue alteration as (for example) delta?
• Will the company have to go back to step 1 (collecting data) each time there is a new variant? At the very least they are going to have to conduct a new trial with each new variant to confirm that the product continues to be predictive.

With the work still to be done I would be very surprised if the company can get a Covid-19 test to market before the December 2022 quarter begins.