[photonic29]

How about cryptogenic stroke?http://www.ncbi.nlm.nih.gov/pubmed/23366107

I think it's relatively clear that the tech works in a controlled setting. It even works surprisingly well in an uncontrolled setting. But as these devices become more mass marketed, sensitivity and selectivity become a much bigger concern. If you detect 99% of all true events, that looks great at first. But how useful is it if only 10% of all detected events are true? Selectivity must be tight before we can issue strong health warnings to masses of people. The challenge of real world noise can be bigger than getting the first principle tech to work.