Kenneth MacCallum, Principal Engineering Physicist, StarFish Medical01.29.19
Ultrasound system development includes some of the biggest electronic, digital logic, and software design challenges of all medical devices. The data generation and throughput can be torrential and the real-time processing requirements ravenous. It's the sort of challenge that will send our electronic engineering (EE) group—myself included—into a tizzy of excitement.
Even as I am drawn to these engineering problems just waiting to be solved, I have noticed it only makes sense under certain circumstances. Nowadays, there are a bunch of vendors that supply original equipment manufacturer (OEM) ultrasound subsystems that can be incorporated into a medical device design. Many unique medical applications can simply be built onto existing ultrasound systems, either purely with software or with the addition of some limited custom hardware. Notwithstanding all of this, there are certain applications that do require a from-scratch solution. Here are some of the conditions that hint such a solution is warranted.
When You Have the Time and Budget
The amount of time and budget to develop an ultrasound system from scratch should not be underestimated. You could argue it involves problems that have been well solved many times. Even though analog front ends (AFEs) are now more integrated and full-functioned than ever before, the printed circuit board (PCB) of an ultrasound system is one of the most challenging and complex designs an EE in this industry is likely to face.
The design will have a substantial layer count with significant signal integrity concerns. A typical 128 channel ultrasound front end can generate over 100 Gbits/s of raw digital radio frequency (RF) data. All of this data must be beamformed down to a lower bandwidth, then squirted over to a GPU for image processing. In addition, the pulsers for all channels must be timed and apodized appropriately, often with nanosecond precision. It should be expected that multiple PCB spins will be required and each spin will be substantial.
Features often overlooked as table-stakes can represent a significant effort to engineer. All signal and image processing software libraries and pipelines will likely need to be written from scratch. The FPGA logic that generates all the timing, marshals the multiple data channels into memory, and pushes them into the data processing pipeline is not insignificant. All of this means a significant investment of time and money that should not be taken lightly.
All in all, if you’re looking to develop a high channel-count, imaging ultrasound subsystem, you should expect it to take many person-years of effort. Unfortunately, this means under most circumstances the conversation transitions to what sort of compromises you are willing to make on your product vision to accommodate an OEM subsystem.
When You Don't Need a Full Ultrasound System
On the bright side, if you don't need a full imaging ultrasound system, there's hope you can have exactly what you want. The fewer features you need, the faster, cheaper, and lower risk the development becomes. For instance, if you only need a few transmit or receive channels, not only is it hardly worth using 128 channel OEM hardware, but the from-scratch development cost could drop by many person-months. If your application does not require full imaging capabilities and does not require sophisticated modalities (e.g. Doppler), this may support the argument to start from scratch.
When You Need Something Quite Different
In certain circumstances, you just can't buy or license what you need. Using an OEM subsystem could be really challenging with significant size and shape constraints. If your application needs something to be super-small or some really funky shape, it's likely an OEM solution does not exist. Also keep in mind these same constraints may greatly increase the challenge to design it from scratch.
If you need a solution that substantially deviates from typical ultrasound system architecture, a from-scratch solution may be all that is left. Maybe your solution involves some novel way of generating ultrasound returns from tissue and therefore needs a totally different circuit from a typical pulser and transmit and receive (tx/rx) switch combo. Maybe you need to operate outside of the typical pulse frequency, pulse repetition frequency, or sample frequency ranges. Sometimes you just can't find an OEM module with the control capabilities you need.
Other times, the available software libraries and APIs do not allow enough flexibility, particularly in the raw RF processing and low-level AFE control. It's almost outrageous to think you might need to design hardware from scratch simply because you need your own custom FPGA logic, but that can be the case.
When You Need a Lower Bill of Materials (BOM) Cost
Some applications may not justify the expense of an OEM ultrasound subsystem. This usually doesn't apply all on its own; you can likely get a multichannel OEM ultrasound system for less than you can build a custom one in low volume. If you combine cost requirements with some of the others, this option may become more compelling. If the capabilities or capacity of an OEM system are beyond the needs of your application, these represent areas of potential cost savings.
Consider Your Ultrasound Design Strategy Carefully
I don't want to rain on anyone's parade if they think they need to design an ultrasound system from scratch. There’s nothing more interesting and fulfilling than this sort of design challenge. Having said that, if you’re going to invest all that time and effort, it really makes sense to consider it carefully and to understand the magnitude of this challenge.
Kenneth MacCallum is a principal engineering physicist at StarFish Medical.
Even as I am drawn to these engineering problems just waiting to be solved, I have noticed it only makes sense under certain circumstances. Nowadays, there are a bunch of vendors that supply original equipment manufacturer (OEM) ultrasound subsystems that can be incorporated into a medical device design. Many unique medical applications can simply be built onto existing ultrasound systems, either purely with software or with the addition of some limited custom hardware. Notwithstanding all of this, there are certain applications that do require a from-scratch solution. Here are some of the conditions that hint such a solution is warranted.
When You Have the Time and Budget
The amount of time and budget to develop an ultrasound system from scratch should not be underestimated. You could argue it involves problems that have been well solved many times. Even though analog front ends (AFEs) are now more integrated and full-functioned than ever before, the printed circuit board (PCB) of an ultrasound system is one of the most challenging and complex designs an EE in this industry is likely to face.
The design will have a substantial layer count with significant signal integrity concerns. A typical 128 channel ultrasound front end can generate over 100 Gbits/s of raw digital radio frequency (RF) data. All of this data must be beamformed down to a lower bandwidth, then squirted over to a GPU for image processing. In addition, the pulsers for all channels must be timed and apodized appropriately, often with nanosecond precision. It should be expected that multiple PCB spins will be required and each spin will be substantial.
Features often overlooked as table-stakes can represent a significant effort to engineer. All signal and image processing software libraries and pipelines will likely need to be written from scratch. The FPGA logic that generates all the timing, marshals the multiple data channels into memory, and pushes them into the data processing pipeline is not insignificant. All of this means a significant investment of time and money that should not be taken lightly.
All in all, if you’re looking to develop a high channel-count, imaging ultrasound subsystem, you should expect it to take many person-years of effort. Unfortunately, this means under most circumstances the conversation transitions to what sort of compromises you are willing to make on your product vision to accommodate an OEM subsystem.
When You Don't Need a Full Ultrasound System
On the bright side, if you don't need a full imaging ultrasound system, there's hope you can have exactly what you want. The fewer features you need, the faster, cheaper, and lower risk the development becomes. For instance, if you only need a few transmit or receive channels, not only is it hardly worth using 128 channel OEM hardware, but the from-scratch development cost could drop by many person-months. If your application does not require full imaging capabilities and does not require sophisticated modalities (e.g. Doppler), this may support the argument to start from scratch.
When You Need Something Quite Different
In certain circumstances, you just can't buy or license what you need. Using an OEM subsystem could be really challenging with significant size and shape constraints. If your application needs something to be super-small or some really funky shape, it's likely an OEM solution does not exist. Also keep in mind these same constraints may greatly increase the challenge to design it from scratch.
If you need a solution that substantially deviates from typical ultrasound system architecture, a from-scratch solution may be all that is left. Maybe your solution involves some novel way of generating ultrasound returns from tissue and therefore needs a totally different circuit from a typical pulser and transmit and receive (tx/rx) switch combo. Maybe you need to operate outside of the typical pulse frequency, pulse repetition frequency, or sample frequency ranges. Sometimes you just can't find an OEM module with the control capabilities you need.
Other times, the available software libraries and APIs do not allow enough flexibility, particularly in the raw RF processing and low-level AFE control. It's almost outrageous to think you might need to design hardware from scratch simply because you need your own custom FPGA logic, but that can be the case.
When You Need a Lower Bill of Materials (BOM) Cost
Some applications may not justify the expense of an OEM ultrasound subsystem. This usually doesn't apply all on its own; you can likely get a multichannel OEM ultrasound system for less than you can build a custom one in low volume. If you combine cost requirements with some of the others, this option may become more compelling. If the capabilities or capacity of an OEM system are beyond the needs of your application, these represent areas of potential cost savings.
Consider Your Ultrasound Design Strategy Carefully
I don't want to rain on anyone's parade if they think they need to design an ultrasound system from scratch. There’s nothing more interesting and fulfilling than this sort of design challenge. Having said that, if you’re going to invest all that time and effort, it really makes sense to consider it carefully and to understand the magnitude of this challenge.
Kenneth MacCallum is a principal engineering physicist at StarFish Medical.