Hacker News new | ask | show | jobs
by dredmorbius 681 days ago
This approach has been tried before. Bill Softky describes a startup he'd worked for using a similar sound-transduction continuous non-invasive blood-pressure monitoring technique. It ... had problems:

Our non-invasive device was supposed to measure blood pressure just as accurately [as an arterial line], but without the cutting, using specially-sculpted sonic vibrations and fancy algorithmic analysis, which was my job. The overall challenge was like measuring the pressure inside a bottle without opening it. Our device worked fine, in that our algorithmically-estimated blood pressure moved up and down, beat to beat, in lockstep with the actual blood pressure. The problem was that our estimate also moved up and down at other times as well, say when the patient moved her fingers, rotated her arm, or took vaso-constricting drugs like nicotine. I spent most of a year understanding these problems, and understanding they couldn’t be solved before our funding ran out. That was when an old-timer taught me an important lesson of measurement: it’s fairly easy to calculate a signal which correlates with what you want to measure, the way our vibration-estimate correlated with actual blood pressure. It’s much harder, though, to calculate a signal which does NOT correlate with what you DON’T want to measure, like arm motion.

<https://www.linkedin.com/pulse/monster-monetization-bill-sof...>

I'd be exceedingly curious as to how the CalTech team have solved that non-correlation problem.

4 comments

It’s just a blurb from a college PR team. As someone that came from academia, they don’t have to have solved any of that because these are generally pretty worthless. It can be as small as someone in a lab discovered how to do the smallest thing and the college wants to run with it to look good.
I'm aware. Which is why I'm pointing to someone who chased that "first" achievement for years, and has the scars to show for it.
There's a nicely-presented introductory paper: https://academic.oup.com/pnasnexus/article/3/7/pgae252/77177...
Paper: https://academic.oup.com/pnasnexus/article/3/7/pgae252/77177...

They tested on the carotid artery. I don't know whether they're concerned with addressing issues of wearing this while active. It seems more likely that it will be used in a clinical setting.

The article mentions getting it down to an arm band or patch.
Looking at the diagram I’d suspect they use accelerometer information from both the wrist watch and the upper arm mounted sensor to remove the effects of arm motion. At simplest it could only check when the arm is in a neutral position. But I’d expect they did something more complex/better than that.
As I understand Softky's work, it's not that the measurements varied predictably with movements, it's that they varied unpredictably.

I'm obviously distant from the project, but a team of SWEs spending years trying to make nondeterministic data deterministic suggests a fairly deep problem.

I guess it's not considered reasonable to keep your arm still when measuring BP, or to measure it only during intervals of minimal movement? Doesn't seem like a showstopper to me.

One of the key applications for this technology is during surgery, when (ideally) nothing is being moved.