Actually "Ultrasounds" are among the cheapest medical devices around. Practically "everybody" makes them and the market is oversaturated.<p>The processing software is actually the least problematic thing; it's all well documented and somebody with the right background (electronics, digital signal processing and computer graphics) could hack it in a single weekend (that's not an exaggeration: When I got invited to the group where I'm currently doing my PhD they were giving me a few datasets of raw, unprocessed Swept-Source OCT fringe data and said "have fun". A day later, using liberal application of NumPy I got pictures; another day and the quality was pretty good.<p>"Ultrasounds" cost more than consumer goods, because, at the moment, they are not consumer goods. Compare this to the cost of "personal computers" in the 1980-ies. Just about as expensive, and "Ultrasounds" are kind of the medical imaging "PC" counterpart for general practitioners.<p>So can "Ultrasounds" be made consumer goods? Difficult, because some parts of them must be built at very high quality standards, not to put the patient at risk. Also some of the electronics involved is challenging, even by todays standards. For example driving the transducers requires driving amplifiers capable of outputting >1kV against a highly complex and poorly matched impedance at bandwidths above 1MHz. That's a really tough problem, that, luckily, has been solved but still requires fairly complex electronics; you can buy appropriate driver amplifier ICs, but those are not cheap, often >10$ per Unit and you need several of them. But that's only half the story: You also need to receive the reflected signal. Here's the problem that the transducers tend to ring after emitting the pulse, causing signal artifacts. And the waves coming back will produce only a few µV of signal. So you've got a 180dB dynamic range between sending and receiving and TX and RX share parts of the signal path; either your RX amplifier can cope with the 1kV sending signal and quickly enough recovers, or you have to add some fairly quick, high insulation signal path switches to quickly switch between TX and RX.<p>And finally you need a whole array of medium speed ADCs (each with a sampling rate of about 10MHz to allow for some oversampling) one for each channel; and of course the interface to the computer. A single 10MHz ADC is cheap. But as soon as we enter the multiple channel interfaces domain things get pricey quick. Just look at audio which operates at most at nimble 96kHz, yet "pro-sumer" (enthusiast consumer) audio interfaces with 16 or more channels go over 1000$; And we need 100 times the sampling rate for ultrasound. So actually the about 3000$ you pay for the ADCs is pretty cheap, if you compare the MHz/$.<p>So you've solved all these essential problems. Now you have to make sure, that a mechanical failure doesn't expose the 1kV driving signal to the transducer to the patient. Here's the challenge: The transducers are separated by the thinnest possible layer of isolation material from the patient's skin, there's a pulsed >1kV amplitude AC signal right behind it, and between the probe and the patient you have conductor gel, which is essentially water jelly, that gives a nice acoustic impedance match, but also does a very good electrical match; we're talking body resistivity model in the two-digit ohms right now. Or in other words: A single manufacturing failure in your scanhead probe and you're going to electrocute the patient. Oh, you're thinking about just floating the whole transducer driver electronics. Smartass, that won't work, because you're operating with MHz AC here, so we're talking RF coupling, driving transducers with significant pulse power; you'll giving your "victim" RF burns, which are nasty.<p>Come to think about it: 8000$ for a ultrasound imaging unit sounds pretty cheap.