Hi, I’m a partner in a Telehealth startup and we’re working on some patents and grants and I need to get the IR of our digital stethoscope (I think!). I want to remove the characteristics of the hardware from the body sound recordings (heart/lungs) so I can “normalize” recording from different devices to more accurately capture the true sound. This will be useful for AI as well doctors listening. We had our device measuredly an audio company and I used that info to create an IR and reverse it (deconvolution I believe is the term). It seems to work well but we need to be accurate as well as measure multiple other stethoscopes. I have a background in music and development (Berklee synth/film scoring dual major and 20+ years in software dev) so I’m familiar with a fair bit of this stuff but I don’t have much experience with scientific side of things.
From what I can tell though the QA403 with an anechoic chamber should work for this but I wanted to confirm and possibly walk through the set and usage. The stethoscope uses a standard usb connection and it just shows up as a mic. I have a UAD Apollo and a Lynx Hilo that I can use to connect to QA but am not sure if that’s the best way to go. I’m also not sure what adaptors I should use to attach everything together, or to be honest, how to run the tests. I assume I can do a sine sweep and/or white noise, record the results, and use that in the python script I wrote but that’s just a guess. Again, I don’t know what I don’t know here.
If anyone has advice or would be interested in helping me walk through this to get me up to speed I’d be happy to pay some consulting fees.
Hi Dan, this is a quick and basic reply about the concept: you first need to find a proper transducer to emit sound that physically couples to your stethoscope like it would to a body. I can’t tell without research what that would look like exactly but it would be some kind of loudspeaker with a flat surface the stethoscope is put in contact with. As the stethoscope uses an USB connection it can’t be used directly with an QA403 or your Apollo or Hilo converters. Easiest (and potentially good enough) way to do it to plug the stethoscope into a computer that runs speaker measurement software (e.g. REW), use the stethoscope as a microphone and send the test signal to an amplifier hooked up the transducer and do your sweep.
Apart from that for production QC you probably will have to come up with an automated version of that to do pass/fail tests and (if necessary) provide calibration files for each unit.
Hi @dangieschen, @THDaniel’s comment makes sense to me. The company you used to measure the mic should have delivered a measurement procedure (or you should have given that to them). Your device is basically a microphone, and usually microphones are tested using a spark gap when precision matters. See link 1) below. For more consumerish applications, you will typically have a “golden” mic with a known response, and compare that to the device-under-test (aka your stethoscope). And then both will be subjected to the same sound (usually pink noise or a chirp). It’s a very challenging problem overall. The most frustrating part of it will be that you could pay Firm 1 to make a measurement, and then Firm 2 to make another measurement, and the agreement between them might not be so good because each will have to make some assumptions that will impact the results.
When things get complicated, the industry players usually come together and create a spec so that you can compare Manufacturer A and Manufacturer B. That spec will get rid of assumptions. The paper @THDaniel refernced published in 2019, and they mention on the first page there’s no standard for specifying how all this could be measured. Maybe something changed between now and then. But in any case, it’s a sizable task you have ahead of you.
Thanks everyone! This is fantastic and a great start to get me thinking in the right way. I’ll ingest all of this and do some testing and see how it goes. Can’t thank you all enough. I was pretty lost.