[tachydyne]

I work for a competitor to Butterfly, and the main difference between this device and other ultrasound transducers is the use of a CMUT array instead of the typical piezoelectric crystal. CMUT arrays are MEMS devices, made in a semiconductor fab using the same processes as other ICs. This makes them cheaper to produce than traditional arrays where the elements have to individually sliced and wired to the electronics. CMUT isn't new, it's been being worked on for 20 years, but it isn't used because it is much less sensitive than the piezoelectric arrays. Either Butterfly has made a breakthrough and are choosing growth over profit margins, or they are trying to make up for worse performance with a much cheaper price. We won't know which until their device is actually on sale.

[QAPereo]

Incredibly informative answer, thanks so much. One last thing, do you personally believe “breakthrough” or “discount” in this case?

[tachydyne]

20 years of CMUT being the next big thing in ultrasound has left me a sceptic, but I know some really smart people at Butterfly. I don't think they would bring a device to market that they didn't think was competitive. My guess is that it'll be okay, not the best imaging handheld device on the market but better than several of our competitors.

[pjwMD]

As an internal medicine phsysician who has been using a handheld daily for the past few years I would only add that high image quality is not always needed for many of our most important applications: Lung, IVC, bladder, DVT, abscess. US is dramatically more accurate than physical exam and stethoscope In these cases.

[tachydyne]

Agreed. I was talking specifically about CMUT vs Piezoelectric. If high image quality was always a requirement, the handheld market probably wouldn't exist at all.

Butterfly is right to focus on getting an answer over image quality. As a startup they will likely be able to iterate quickly on machine learning, which is critically important to growing the ultrasound market. Right now the biggest barrier to adoption is the difficulty of reading ultrasound images, and machine learning is the most promising solution to that. I think that is more likely to be their competitive differentiator than their use of CMUT.