[CamperBob2]

Not OP, but FPGAs are good for realtime signal processing, including initial processing needed to filter and downconvert high frequency signals to make them accessible to conventional CPUs.

They are also parallel processing-oriented by nature, so a traditional ultrasound implementation based on an array of sensors is a good fit for an FPGA-based front end.

People tend to use FPGAs for tasks that used to be offloaded to dedicated DSP chips. Besides their processing capabilities, one key reason is that their I/O capabilities can handle just about any data acquisition/interfacing problem. When you need to move a lot of data around in a hurry, there are few solutions better than an FPGA, and certainly no more flexible ones. For mass production, of course, both DSP chips and ASICs tend to be more economical.

Some of the tasks involved in ultrasound could potentially be handled by the PRU units on the BeagleBone Black platform. Not sure if that would work well for this application but could be worth considering.