Application Menu

Medical Case Study

GPU Computing Solves Processing Challenges of 4D Ultrasound

The Situation

The modern medical office works at breakneck speed with back to back appointments. For new technologies to be effective in this environment they need to be able to keep up with the pace. Results need to be turned around in minutes, not hours, to keep patients happy and offices running smoothly. As ultrasound technology has progressed, new forms of detailed 3D and even “4D” moving ultrasound video have been made possible resulting in better diagnosis, but could take hours of image processing on multiple server-class computers -- until now.


The Challenge

3D ultrasound imaging is extremely computationally intensive and to perform it in real time, necessary for the latest 4D ultrasound, requires computing power beyond that of any general processor. 

Besides powerful signal processing capability, the customer had a host of diagnostic features they wanted to include which would increase the value of their product to physicians, but limited development time, so they wanted the ability to rapidly prototype and test various imaging algorithms

As a medical equipment provider, the company also wanted a partner that would not only solve its compute performance in the short run, but provide long term support through a multi year development and service life cycle. 

The Solution 

Aetina was looked on favorably for its experience in the GPU solution industry and the ability to offer a high performance graphics product in a form factor that integrated well with the company’s proprietary COM Express based ultrasound system.

Whereas the manufacturer was originally wary of GPU technology because of the short term support and fast product turnover of commercial providers, Aetina’s ability to work with them through their multi-year development process with hardware and firmware customization as well as promise product longevity support finally won them over.

We listened closely to our customer and worked closely with them through the entire development process to understand their ultrasound equipment’s demands on computational performance, as well as providing guidance for CUDA applied development. Starting from the initial product conceptualization to prototyping, optimization, through five generations of specification changes, our engineering team worked in mutual collaboration to make sure the proper hardware was configured and that firmware was optimized correctly for ultrasound performance and for the value-add image processing features the customer desired.

In the end the Aetina M3N970M-MN chosen, a MXM 3.1 form factor graphics card featuring NVIDIA GTX 970M graphics processing unit with staggering 1280 CUDA cores. This compact GPGPU card was adapted via MXM carrier on a COM Express board to fit with the manufacturer’s proprietary embedded system. 

The Results

The M3N970M-MN completely met the high performance computing needs of our client’s 4D ultrasound equipment. The 1280 CUDA cores provided more than enough computing power to perform advanced 3D visualization of ultrasound data in real time using the latest phase shift harmonic imaging. Combining the advantage of its software based CUDA platform for rapid prototyping, and the extraordinary computing power of GPGPU computing, allowed the customer to provide additional image processing features including echo-enriched beam forming, structure detection, curved MPR, and much more, helping the customer distinguish their equipment from competitors, as well as forming product segmentation. 

Thanks to Aetina’s service commitment and technical support the customer enjoyed close collaboration throughout their development process with guidance on CUDA based development. With the upgradability of CUDA based computing and our product longevity support the ultrasound company could be confident that its equipment will have a long service life.

Related Products