Ai Detects Covid

[naomi24]

An app that detects Covid in your voice has been developed, and according to scientists, is easier to use and more accurate than a lateral flow test. The Dutch researchers say coronavirus usually affects the upper respiratory tract and vocal chords, leading to changes in a person’s voice, but I wonder how is distinguishes between changes in the voice caused by colds, 'flu, or laryngitis, say, and those caused specifically by Covid? Any ideas?

https://metro.co.uk/2022/09/05/app-that-detects-covid-in-your-voice-more-accurate-than-lateral-flow-17299441/

[brian j john]

Would it work if i spoke Welsh into it

[naomi24]

The Dutch developed it, but Welsh is a thought. Double Dutch perhaps? ;o)

[brian j john]

Da iawn ,,

[TheChair]

Surely it can only detect a change in your voice if it already has a sample of it when you are covid-free?

[naomi24]

Good point.

[jim360]

I'm completely clueless about how such things work. All I can say is that it's remarkable what they seem able to achieve these days by throwing a neural net at the problem!

[Peter Pedant]

Oh I thought Jim would give us a bit on Machine Learning
(alot on U tube on ML which I find terribly boring)

Basically - your normal voice is like Audrey H - in SPAIN ! in SPAIN . And when you have covid AND layngitis ( hem hem) you sound like Ermintrude Magic Roundabout - hello dahling....

and so this is a machine that spits out YES or NO
( the problem is decidable - watch out Jargon word alert)

[Peter Pedant]

clearly, no laryngitis ( and I cannot remember how many present with laryngitis) - no pos covid
BUT
They have used a training set ( 5000 persons) and of those 800 were covid positive

and they were quite within their rights to test it against the next best ( or one of the best) lat flow tests.

https://www-users.york.ac.uk/~mb55/meas/ba.pdf

Bland Altman comparison - cited by eggheads 52 000 times

[Peter Pedant]

and you customarily do it so that your thing looks very good indeed.

BUT ... they quote sensitivity and specificity ( jargon alert) which applies to the test and what you want is....
how good is it in the population
for which the statistic 'positive predictive value' is the asnwer to the question
I have a positive test, what is the chance I hve COVID

this involves Sp and Se AND how common the disease is ( prevalence - nope not incidence - anyone remember the long long arguments whether or not prevalence per unit time is incidence? it is, by the way)
and I cant see a PPV - so the easy way of telling if a test is any good
I have tested pos, how likely is it I have covid

is not answered - - - - jury still out

[Sqad]

No idea at all.
The OP is correct in questioning the differential diagnosis.

[dave50]

What is the point? Time get over it and move on, some may still get covid a few may even die but hey ho that's life.

[Peter Pedant]

the tee-vee doc said
this will catch on if it works - no one slike sticking a swizzle stick up their schnozzle

I try to educate AND use the vernacular recognised by the averagge ABer

[Peter Pedant]

so how does it disctinguish between Covid and plague?
er well
in plague you lie down and get to be dead very quickly

so it has to be in a high prevalence ( of covid) area
( so NOT christmas island, or Orkney) and the training set was.

[dave50]

Great idea, hyperchondriacs roaming shops and bars with the app on their phone, trying to find those with the dreaded lurgy.