Read on Twitter
Here's some interesting finding of published article regarding "electronic nose" screening method.
since apparently no scientific publication whatsoever have been released by the genose researcher for the - already widely used - device.
since apparently no scientific publication whatsoever have been released by the genose researcher for the - already widely used - device.
Interesting start:
weird, abstract only include sensitivity and NPV.
reasonable since the purpose to screening. but still, 100% sensitivity with close to 50% specificity is not a good option. even for screening.
reasonable since the purpose to screening. but still, 100% sensitivity with close to 50% specificity is not a good option. even for screening.
The readiness of this screening is still troubling when you see the area under the curve.
suppose you make cutoff point where the sensitivity is 92%. (Y axis) that would means the chosen specificity is around 50-60% (1-X axis)
suppose you make cutoff point where the sensitivity is 92%. (Y axis) that would means the chosen specificity is around 50-60% (1-X axis)
what's that translated into?
it means that around 50% of actually negative will appear as positive.
that would put a load of burden to confirmative diagnostic method such as PCR.
it means that around 50% of actually negative will appear as positive.
that would put a load of burden to confirmative diagnostic method such as PCR.
Say in population of 1000, 100 people infected.
that means this screening test with 90% sensitivity, and 50% specificity will show (90 (true positive) + 450 (false positive) ) 540 positive results.
while the real positive is just 100 people.
that means this screening test with 90% sensitivity, and 50% specificity will show (90 (true positive) + 450 (false positive) ) 540 positive results.
while the real positive is just 100 people.
how to make sense of how messy is this thing is through a scattered plot.
to see how much the test intersect between real positive and real negative patients. with green dots being a real positive patient.
positive results is any dot above the line.
to see how much the test intersect between real positive and real negative patients. with green dots being a real positive patient.
positive results is any dot above the line.
Is the plot looks like the method is ready to be applied to you?
idk, but i doubt it for now.
but genose manufacturer said that it AI based system which means it can be improved overtime. and we have to wait the real publication from the researcher to know for sure.
idk, but i doubt it for now.
but genose manufacturer said that it AI based system which means it can be improved overtime. and we have to wait the real publication from the researcher to know for sure.
but it's very disappointing to see the attitude of such researcher on that interview. any researcher usually loved to talk about their results, and tried as hard as they can to make people understand
