[atria]

Disclaimer: I'm writing software for a company that offers FDA approved pace-makers, defibrillators, and implantable heart monitors. These comments are my own.

No where on your site do you reference FDA approval and I see that you are not in the U.S., so take my comments with a grain of salt.

Bio-metric authentication using cardiac wave forms? I'm not sure that is a good way to go, but it sure sounds good. What happens when the user takes medications that changes the waveform (ie., they start new blood pressure medicine or cardiac meds, or intermittently take those meds)? What if the user has PVCs when trying to authenticate? What about diabetics with high glucose levels?

The only time I've seen a legitimate use for using ECG in a consumer/commercial application is detecting a pulse when a user is attempting validate a finger print. I worked with a guy who had to implement something for ATMs being shipped to Africa. They required the user to authenticate with a finger print scanner in order to withdraw money. It wasn't long before people were loosing their fingers.

As far as monitoring for cardiac anomalies, they don't happen all the time. Your device only appears to work when you hands are touching it. It often takes days or weeks for them to occur, requiring continuous monitoring for days, weeks, and sometimes months.

[capcarr]

Medical certification is something that we're not pursuing at the moment. For now, we intend to raise alerts to the user, in case we detect something suspicious, referring them to their medical doctor. We do not claim to detect cardiac pathologies that are rare, and need several ECG derivations to be correctly diagnosed.

What we do offer is the possibility of more frequent ECG monitoring, beyond the 10 second ECG exam that most people do only once or twice per year. For example, driving is a stressful environment, where a major cardiac event is more likely to occur. In a car with our technology, we would be able to detect when that event happens, thus preventing a crash by automatically stopping the car, and calling emergency services.

Regarding signal variability due to medication and other pathologies, that's something we have to account for, with robust feature extraction methods that are invariant to these changes, or adapt to them.

[atria]

I'm not familiar with European regulations, or medical procedures, but in the U.S., a medical diagnosis can only be made by a qualified licensed medical professional. Even making statements about diagnosing disease states via software or smartphones is enough to invoke an FDA response.

[tgokh]

This is hugely important. EKG changes are common - drugs change the QT interval, heart disease and subacute ischemia can cause T wave changes, people can go in and out of atrial fibrillation or throw PVCs, etc. I'd be very impressed if you can reliably identify a user over a period of years using EKG traces.

Your use of EKG to determine "in-game focus", "driver fatigue" and workout quality is definitely interesting. Is there literature showing reliability of these measures from EKG readings and heart rate, or are you using other sensors in addition to EKG? You mentioned in another reply that your group has published papers on EKG biometrics - you should definitely link to or at least cite those on your website.

[capcarr]

We'll post some links, thanks for the suggestion.