Embedded AI

Offenburg, July 11, 2024

“The magic lies in the unused data,” says Viacheslav Gromov, CEO of AI developer AITAD. This article deals with the question of how embedded AI can be optimally integrated into medical devices.

What should medical device developers pay particular attention to when integrating embedded AI?

In addition to the benefits for doctors and patients, what is needed, alongside the big vision, are, above all, practical development steps. Proof of concepts for individual system components, such as intelligent embedded AI sensors, can be a good starting point and also ensure the necessary acceptance among management.

What is often forgotten when selecting components based on performance: Especially in the case of battery-powered medical devices with an NPU peripheral unit, energy consumption should not be neglected despite high efficiency and must be taken into account in the design.

And last but not least: For the basic development, companies should initially free themselves from a healthy portion of the regulatory burden. AI is currently developing so quickly that regulation is lagging behind with a certain latency.

“The magic lies in the unused data,” says Viacheslav Gromov, CEO of AI developer Aitad. In this interview, he provides an overview of current AI trends, hurdles such as the AI Act, and helpful AI tools for developers of intelligent medical technology.

Mr. Gromov, from mega chips to GenKI – which AI developments are currently most relevant for medical devices?

It is clear that processors with high computing power and generative artificial intelligence are accelerating both medical research and the development of medical devices. This year in particular, global AI successes have generated more acceptance and interest in the economy and society than ever before. At the same time, the fundamentals – especially in the embedded sector – are becoming more and more performant and marketable, with AI-optimized semiconductors such as the recently introduced NXP i.MX93 with the Arm Ethos U65 NPU or the Renesas RZ/V2H with its own DRP-AI3 accelerator, to name just the flagships.

The typical technical and non-technical technology curves are therefore currently converging – which makes artificial intelligence in products and especially in medical technology so exciting and tangible as never before due to special safety and efficiency requirements.

What should medical device developers pay particular attention to when integrating embedded AI?

In addition to the benefits for doctors and patients, what is needed, besides the big vision, are, above all, practical development steps. Proof of concepts for individual system components, such as intelligent embedded AI sensors, can be a good starting point and also ensure the necessary acceptance among management.

What is often forgotten when selecting components based on performance: especially in the case of battery-powered medical devices with an NPU peripheral unit, energy consumption should not be neglected despite high efficiency and must be taken into account in the design.

And last but not least: for the basic development, companies should initially free themselves from a healthy portion of the regulatory burden. AI is currently developing so rapidly that the regulatory system is lagging behind with a certain latency.

Read the full interview at elektroniknet.de (in German)