Here's what it takes to make IoT data ready for AI and machine learning

Hospitals should begin with appropriate infrastructure and that includes robust connectivity, storage and security. Here’s a look at how to get started.
By Bill Siwicki
11:57 AM
Share
IoT data ready for AI

AI is poised to gain traction but still needs better integration with EHR systems and analytic tools for medical devices in the hands of patients.

The integration of artificial intelligence and the Internet of Things introduces a wide array of connected health tools that produce a vast amount of data that must be synthesized, analyzed, stored and communicated by a robust information infrastructure. But if hospitals don’t structure and store IoT patient data properly, that information could be rendered not assessable by AI tools.

For starters, significant infrastructure is needed to streamline IoT-generated data to make sure it is simple to assess and manage with AI. 

“AI adoption and scale will be accelerated by the relatively low cost of deployment,” said Rick Krohn, president of HealthSense, a connected health consulting firm. “A terabyte of storage costs less than $100, and wearable sensors and cloud infrastructure are becoming increasingly affordable. But AI requires sophisticated applications that deliver contextually aware right-place-right-time clinical decision support.”

[Also: How AI is transforming healthcare and solving problems in 2017]

Indeed, healthcare organizations of all types must ensure that their networks are secure enough to absorb and manage multiple devices with their attendant bandwidth and storage requirements, he added.

“Key baseline requirements around IoT-ready infrastructure include the ability to handle real-time data streams, and the capability to horizontally scale given that there will be increased volume, velocity and veracity of data and applicable storage services,” said Gary Palgon, vice president of healthcare at data management technology company Liaison Technologies, whose offerings include AI. “There are many other requirements needed to later synthesize the data, such as resident and alternate memory functionality, but those are depending upon the business needs.”

[Also: AI vs. COPD: Organizations bet on machine learning to improve lives]

Solutions to IoT infrastructure challenges exist today, but primarily only address niche use cases and are not available to scale across the myriad of IoT scenarios that are being tested or used in healthcare today, he added.

“The infrastructure challenge is to be able to have a single platform that addresses the Big Data needs while being broad enough to handle the variety of use-cases being developed around IoT, with data as the central focus,” he said.

So what steps can healthcare provider organizations take today to make sure their IoT-generated data is ready for a robust and helpful application tomorrow? There is a variety, in fact, and healthcare CIOs would be wise to consider them amid the growing influence of both the IoT and AI.

First and foremost, the IoT data that is currently available needs to be captured, ideally with context, but until data sources are able to handle that, without context is fine, too, Palgon said.

“While there are many standards that exist or are being developed, standards will not solve the data readiness problem, as we often don’t know in advance how data will be interpreted,” Palgon explained. “If it can be made available in a structured format, that is helpful, but in the end, there are many new forms of data being collected, including semi-structured and unstructured, including images, so just the ability to capture it is core.”

Storing the information in a Big Data environment, sometimes called a data lake, is a useful step, but many sources will be located in distributed environments, so making those digitally accessible is also helpful toward allowing the information to be consumed at a later date.

“The value of data goes up when it can be blended together to create new insights,” he said. “So moving away from application-centric models where data is bound within applications in pre-structured relational databases and moving to open, data-centric environments makes it helpful for interpretation.”

Krohn said that in the near term AI will gain traction as a tool to cut costs, address readmissions and refine clinical decision support, but that is going to require the integration of AI with EHR systems and analytic tools.

“To align with AI systems, IT investments should have robust storage, security and connectivity capabilities,” Krohn said. “Operationally, healthcare providers should recognize that the convergence of the IoT and AI will impact the way healthcare occurs. Smart systems will allow provider organizations to predict health issues and events, to prevent and intervene earlier and more effectively, and to more capably serve those consumers in greatest need of longitudinal, intuitive tracking – our growing chronic and aging populations.”

Twitter: @SiwickiHealthIT
Email the writer: bill.siwicki@himssmedia.com