As the smart hospitals trend continues to grow, medical technology developers are under increasing pressure to integrate AI into their devices. This means those technology developers that fail to make the shift to AI-powered medical solutions could be left behind.

The problem is that lengthy and expensive development cycles have been an obstacle to integrating AI into medical devices. Besides stringent certification requirements, it’s complicated and time-consuming to choose a machine learning framework and deploy an AI application on compatible hardware. AI technologies are also evolving at such a rapid pace that it’s hard for technology developers to bring new products to market before they become obsolete.

In this blog, we’ll provide an overview of the steps involved in developing an AI-powered medical device from scratch. We’ll also discuss how using hardware and software building blocks can reduce time-to-market by 7x.

Developing an AI-Powered Device from Scratch

There are two key aspects to developing an AI-enabled device: the hardware — including the components that power the device and the housing that encloses them — as well as the software that implements machine learning.

The Hardware Manufacturing Process

Developing a custom medical hardware solution from scratch can take up to two years. Here’s how the typical manufacturing process looks:

• Design: Industrial, mechanical, and electrical design for custom devices takes months because components need to be deliberately researched and selected based on cost, compliance, compatibility, and many other factors.
• Configuration: Putting together AI components and configuring different peripherals, such as keyboards, monitors, and articulating arms, can be time-consuming because everything has different setup requirements.
• Testing: Troubleshooting and addressing manufacturing and quality issues can take several more months, depending on the complexity of the device.
• Certification: Obtaining baseline certifications like IEC 60601 takes a few months, plus any additional certifications for the healthcare industry.
• Ramping up Production: Once the final device is ready, it can take time to ramp up production for the new product before it can go to market.

Building AI-Driven Applications

It can also take software teams up to 80 weeks to have a usable AI application when developing machine learning models from scratch. Here’s some of the steps involved:

• Collecting and labeling data to feed the machine learning algorithm can take months.
• Training the model can take a lot of time, depending on the machine learning platform.
• Tuning and pruning to achieve performance targets often takes several more months.
• Deploying the model into a machine learning pipeline can take several weeks.

Reducing Development Time with Modular Hardware & Predefined Software

MBX’s innovative building blocks approach can eliminate much of the legwork leading up to the last mile of development. That means technology developers can use modular hardware and predefined software as a starting point, and then tailor the solution to fit their unique requirements.

Hardware building blocks can reduce the typical development from two years to just four months, which is 6x faster than starting from scratch. Rather than spending months designing a fully-custom device, technology developers can choose from hardware building blocks that are pre-tested, pre-certified, and optimized for AI use cases. After branding and customizing the device to meet specific requirements, the existing production lines of the base building block can help deliver the new product to market faster.

Similarly, medical technology developers can achieve a 10x time savings by using predefined software like NVIDIA Clara Guardian. This machine learning platform provides pre-trained models, an optimized SDK for training and inference, and many other capabilities. Altogether, this can help the software team deliver a working AI application in a matter of months rather than years.

Delivering a Mobile Computer Vision Medical Device 7x Faster

MBX offers numerous hardware building for computer vision solutions, embedded AI systems, and server-based AI solutions. In addition, MBX has partnered with NVIDIA to accelerate AI device development in healthcare, making it possible to deliver a computer vision medical device in just six months.

More specifically, MBX Kori is a modular mobile AI platform that is optimized for computer vision applications using NVIDIA Clara Guardian. Together these hardware and software building blocks can help technology developers deliver a mobile computer vision medical device 7x faster than starting from scratch.

Delivering an AI-powered medical device doesn’t have to be a long and expensive process. By partnering with MBX, medical technology developers can have the guidance they need to develop complex solutions in a fraction of the time.

Learn more about Fast-Tracking the Development of AI Workflows in Hospitals with MBX and NVIDIA.