MathWorks accelerates development of software-defined workflows in medtech with NVIDIA

Developers can use their existing MATLAB algorithm codebase to develop and deploy AI-enabled applications using the NVIDIA Holoscan platform.

With GPU Coder, engineers can automatically generate optimized CUDA code from MATLAB functions, allowing them to build custom operators within the Holoscan SDK. These operators can then be integrated into software-defined, hardware-accelerated, low-latency medical applications that handle real-time streaming I/O data.
With GPU Coder, engineers can automatically generate optimized CUDA code from MATLAB functions, allowing them to build custom operators within the Holoscan SDK. These operators can then be integrated into software-defined, hardware-accelerated, low-latency medical applications that handle real-time streaming I/O data.
CREDIT: BUSINESS WIRE

MathWorks, the leading developer of mathematical computing software, announced an integration that makes MATLAB available in the NVIDIA Holoscan platform. Medical device engineers can now accelerate the development and deployment of streaming data analysis and visualization applications by wrapping existing MATLAB algorithms and functions into GPU-accelerated NVIDIA Holoscan operators for real-time data processing and inference.

 

Medical device engineers must navigate complex, ever-evolving global regulations while keeping pace with rapid technological innovation using cutting-edge materials and electronics. These conflicting forces cause many devices to be near-obsolete shortly after reaching the market and fueled the emergence of Software as a Medical Device (SaMD). The International Medical Device Regulators Forum defines SaMD as “software intended to be used for one or more medical purposes that perform those purposes without being part of a hardware medical device.” To stay competitive in the market after deployment, SaMD products require the capability to integrate additional software feature functionality over time, necessitating engineers to develop a software-defined workflow for their development.

 

“The medical technology industry is being transformed by artificial intelligence,” says David Niewolny, director of business development for medtech at NVIDIA. “By providing a seamless development experience for the growing MATLAB development community within NVIDIA’s medical-grade Holoscan platform, NVIDIA and MathWorks are accelerating AI-powered innovation in med tech.”

 

NVIDIA Holoscan is a sensor processing platform that streamlines the development and deployment of AI and high-performance computing applications for real-time insights. The platform helps bring the latest AI applications into clinical settings by providing full-stack infrastructure for scalable, software-defined streaming data processing at the edge. The MATLAB integration into Holoscan will help medical device engineers use existing built-in matrix operations and complex toolbox functions for image and signal processing, filtering, transforms, and deep learning algorithms. Implementing a Holoscan pipeline with MATLAB requires four steps: creating a MATLAB function, generating accelerated CUDA code using GPU Coder, creating a Holoscan operator wrapper, and re-building the Holoscan application with a new MATLAB operator. This allows the creation of a completely functional, software-defined workflow for medical devices. Additional integrated verification and validation capabilities in MATLAB and Holoscan allow developed software-defined workflows to comply with industry regulations and standards, including IEC 62304.

 

“Engineers can now write MATLAB functions and execute them thousands of times faster with NVIDIA Holoscan,” says MathWorks’ Director of Product Marketing David Rich. “Our collaborations with industry leaders like NVIDIA are set to drive medical device innovation as millions of our customers seek to design, develop, and test their products while complying with industry regulations and standards.”

 

More information on the benefits of the MATLAB integration into Holoscan can be found at the following links: https://github.com/nvidia-holoscan/holohub/tree/main/applications/matlab_gpu_coder