[picklefish]

“The biggest problem in detection of such cells is that these cells are extremely rare- one in a billion blood cells,” says Goda.

According to Goda, the cutting edge device now harbors an unprecedented false-positive rate of one cell in a million.

For anyone who has forgotten statistics like me and was thrown off by this: https://en.wikipedia.org/wiki/Type_I_and_type_II_errors#Fals...

The false positive rate is the proportion of absent events that yield positive test outcomes, i.e., the conditional probability of a positive test result given an absent event.

[bergerj]

1 of 1B - [pre-cancerous cell] to [normal cell]

1 of 1M - [false-positive cell] to [cells tested]

So, given my poor stats background and guessing what that means...

1B - cells tested

1k - false-positive cells

1 - pre-cancerous cell

-->

20T - ALL cells tested

20M - false-postive cells

20k - pre-cancerous cells

http://en.wikipedia.org/wiki/Red_blood_cell#Human_erythrocyt...

[Ralith]

So one in one thousand positive results is legitimate. How is this tool remotely useful?

[DanBC]

They don't mention their false negative rate, nor any of the rates for competing tech. So we can't say that this device catches the one in X that other machines don't catch. And we can't say that this machine finds 1 in 1,000, where competing machines only finds 1 in 100,000.

In theory early detection is good, but I'm not sure what treatment is available if cells are found this early ad the stress for the other 1,000 people is going to be significant.

[Ralith]

So one in a million healthy blood cells will register as cancer, while the expected frequency of true positives is around one in a billion. Maybe I'm still screwing up the statistics, but doesn't that mean that the results of testing cancerous and normal blood will be almost indistinguishable?