Building Foundations to Build Better Care

It’s all about the data. We’ve been saying this for years. We can choose to look at this in one of two ways. It’s either a constant truism or it actually evolves and gains mass over time. In the age of artificial intelligence, it is both. 

In the last issue of this magazine, one small subhead offered great truth: “Don’t let data access derail clinical breakthroughs.” We can’t, and thus need to build a solid strategy so that can’t happen. 

Healthcare is on a mission to innovate, powered by artificial intelligence and machine learning. Collecting data, storing data, learning from data—and even more so, grabbing a rear-view mirror to peer back into data to see what may have been missed. And doing that again and again over time. The value of data are now, and later, and thus always need to be nearby. That is what physicians, data scientists, engineers and informaticists need to generate breakthroughs and good outcomes on a daily basis.

The way forward is supported by smarter infrastructure that evolves along with data needs. At its core is a data hub with high computing power and a massively parallel structure, storing and serving up data quickly. We call this storage technology FlashBlade but this is why it matters. It offers a much higher rate of access from a much larger pool of data. In moving from rotational spinning disk to flash, enterprises are no longer bound to the latency principles of spinning disk or the throughput limitations of traditional disk systems. Hospitals, health systems and AI developers already using it are succeeding in accelerating and achieving innovation. To put it more simply: Storage got better, compute got faster, and now physicians and researchers are going deeper into these massive data stores to find more robust results. 

Let’s look at this from the perspective of a pediatric radiologist. Their numbers are low but the need for their skills is high. So there would be great benefit to developing a machine learning application, for example, that looks at 100,000 scans to detect anomalies—significantly greater than a person could ever accomplish—and identify 421 with anomalies that need to be viewed by a human. This would be significant, and accomplished in seconds.