New AI-enabled technology marks real advancements for medical edge and embedded AI applications

Artificial intelligence at the edge has inspired a wave of new ideas and applications that were not even possible when RSNA was last held in person just two years ago. Chips are getting faster and cheaper, with AI processing performance jumping as much as 14x per generation. In addition, SDMs and parallel computing software platforms like CUDA allow software to use certain types of GPUs for general purpose processing. This has made a huge impact on advancing radiology and imaging, with the application of AI and computer vision in the healthcare field becoming the gold standard.

Developing AI technology can be daunting. Inferencing is complicated, with applications having different requirements, use cases and operating environments. It can take upwards of 80 weeks to develop a model from scratch, which consists of collecting and labeling the images, and then tuning and pruning the AI model.

Another challenge is making AI solutions ubiquitous, using pre-designed software and hardware “building blocks” as the basis to compress development time and bring solutions to market faster rather than one-off designs for every application. A lot of headway has been made here too, since RSNA 2019.

Building Blocks for 10X Faster Development Time

One of the benefits of building blocks is that they help developers get a head start at developing the AI models with a solid underlying architecture. Leading the charge with software building blocks is NVIDIA Clara. This healthcare application framework for AI-powered medical imaging provides pre-trained models with training toolkits tailored to medical inferencing solutions. It has effectively made artificial intelligence applications much easier to deploy.

On the hardware side, MBX has developed a series of building blocks to help developers efficiently identify and deploy the physical solution to support AI. Hardware building blocks are reference platforms specifically designed to optimize AI applications deploying as embedded systems or edge devices or servers.

Working in tandem, MBX’s hardware building blocks and NVIDIA Clara software simplify and accelerate the way medical AI, machine learning and computer vision applications are brought to market.

Take a look at a few of the new artificial intelligence hardware platforms MBX is introducing in time for RSNA:

Varion® Compact AI-Enabled Embedded Systems

These pre-configured embedded systems have been thoroughly tested and certified for medical deployments, allowing you to bring AI inference applications to market faster, easier and at a lower cost. Each platform takes advantage of NVIDIA Jetson Xavier architecture and is designed to enable inferencing on the device.

Varion G1-PSF-JNX

Supports deep learning trained models
High performance with lower power consumption than a full server
NVIDIA Jetson Xavier NX
NVIDIA JetPack SDK
Compact fanless design
Wide operating temperature range
Supports cloud native applications

Varion G1-PSF-JAGX

High performance with lower power consumption than a full server
NVIDIA Jetson AGX Xavier
NVIDIA JetPack SDK
Compact fanless design
Bundled with Linux Ubuntu 18.04
Supports a PCIe Add-on Card
Supports cloud native applications

Varion High Performance AI Edge Server

In the world of AI and machine learning, the benefits of high-end processing capabilities near the sensor devices provide optimal results. With the new Varion AI Edge Servers, you can securely manage and scale AI deployments across unlimited edge devices.

Varion G1-P2R8-ID

Ideal for diagnostics, imaging and similar medical use cases
Supports multiple AI inputs or cameras for inferencing and analytics requirements
Enables on-device inferencing to increase performance and reduce time for processing AI analytics
Multi-GPU rackmount AI server featuring 1-4x high performance NVIDIA A30 GPUs
2U compact rackmount unit locatable in a central data center or tuck into its environment
NVIDIA Clara software suite compatible

Varion P2

Ideal for patient monitoring, diagnostics, imaging and similar medical use cases
Full height double-wide GPU support in a compact footprint
Functions as a smaller AI edge device or embedded platform
Powerful for embedded applications needing on-device inferencing
Optional VESA mount enables mounting of the unit to many different devices
Single PCIe x16 slots can accommodate a variety of cards

It’s clear that AI technology moves incredibly fast, and MBX can help you move faster! These new platforms are the tip of the iceberg when it comes to quick-to-market AI solutions. View our other AI/inferencing embedded and mobile platforms and various other medtech architectures and learn more about our custom design engineering and manufacturing services on our website. And contact us if you’re interested in speaking with the hardware and software engineers from MBX and NVIDIA at RSNA 2021.

About Roger Lam

Vice President of Engineering

Roger leads a multi-functional team of engineers at AHEAD that develop, verify and maintain next-gen hardware platforms for complex technologies including Kori, AHEAD’s modular mobile platform for computer vision applications. Roger has 18 years of experience developing medtech computing devices for developers of all sizes, including GE, Siemens and Phillips.