Medical imaging is improving with the use of artificial intelligence (AI)

I’ve been reflecting on this GTC 2021 session discussing how AI is affecting medical imaging, and of course the edge AI solutions to help address industry changes. It’s clear that edge computing and the GPU, together, are playing a key role so I wanted to briefly share what we’ve been seeing, and two of my favorite features enabling AI in medical imaging technology.  

I know I’m not surprising anyone by saying the cost of AI hardware, and software for that matter, is going to skyrocket in no time. However, I feel it’s necessary to reiterate this point in order to place a keystone for the rest of this blog. In a recent market analysis, the global medical imaging market is expected to grow at a compound annual growth rate of 5.2% from 2021 to 2028, reaching USD 28.6 billion by 2028. While many medical applications using AI will see increased growth compared to others, medical imaging is seeing quite the surge. 

Medical Imaging, GPUs and Edge Computing  

The newest hardware offers a lot when it comes to the performance needed to solve problems such as responsiveness, the need for better accuracy, and enhanced imaging. Fortunately, GPU computing improves medical imaging for healthcare professionals, so they can make quicker and more accurate decisions. 

We clearly see the industries’ revolution in terms of medical imaging technology. This is due, in a big way, to NVIDIA‘s advances in GPU technology which complements ADLINK advances in edge AI hardware and software. Edge computing, with the addition of a GPU, improves image quality, enhances the process of image construction, and completes image analysis to assist medical professionals. Intelligent video analytics (IVA) is a huge advantage to GPU solutions for both clinical practices and surgical procedures. 

The embedded mobile PCI express modules (MXMs) is another GPU standard that also enhances medical imaging applications. The combination of edge computing with GPUs in medical imaging provides the powerful computing workloads image reconstruction requires along with the massive amount of heat generated. In applications such as clinical examination, biopsy, intraoperative monitoring and even surgical treatment, the performance-per-watt and even extended temperature options in the computing environment can dictate the speed and accuracy of image reconstructions.  

Features that Make a Difference: MXMs, RDMA  

MXMs 

A wide range of processors can be deployed to suit specific applications. ADLINK’s embedded MXM GPU modules have the ability to enhance performance beyond the traditional CPU. 

Our MXM modules are developed with the NVIDIA Turing architecture, which integrates CUDA cores, RT cores, and Tensor cores in one GPU. One example is the EGX-MXM-RTX3000 module which features advanced NVIDIA Turing GPU technology in an MXM 3.1 Type B form factor. This is just a quarter of the size of full PCI Express graphics (PEG) cards. The modules, which start with a power of 80W are suitable for size, weight and power-constrained (SWaP), critical medical imaging applications. 

Applications in the medical space also crave the longevity of applications in the industrial spaces, which is achievable with joint ADLINK and NVIDIA soIutions.   

RDMA  

NVIDIA’s GPUDirect Remote DMA (RDMA) is a “family of technologies that enhances data movement and access for NVIDIA data center GPUs”. Implementing RDMA provides external data sources with direct access to the GPU’s memory. Without RDMA in medical imaging, the data would go into a CPU’s memory, which may cause further delays in data transmission and latency.  

One use case that involves RDMA, for examples, is ultrasound imaging. Ultrasound leverages front-end devices such as FPGA for analog-digital conversions before data reaches its final destination—the GPU. This presents massive amounts of communication between the FPGA and the GPU. RDMA allows the bandwidth to increase, which provides the computing performance needed for ultrasounds. 

Missed the Session?  

And there is so much more! Hopefully we’ll be back in person again soon, otherwise, if you missed the original session, you can learn more about ADLINK’s medical-imaging technologies in this talk or by visiting us here

Author: Zane Tsai
Author: Zane Tsai

Director of Platform Product Center, Embedded Platforms & Modules, ADLINK Technology

Leave a Reply

Your email address will not be published.