Deployment and scaling issues are the real obstacles preventing healthcare AI from generating value, according to AI company executives. Nvidia And Hoppr.
This is why they are moving away from creating stand-alone models and focusing on the infrastructure needed for these models to actually be used in clinical practice. Hoppr has built an AI foundry that uses Nvidia’s computing and core models — an offering, partners say, that gives developers access to tools to more easily launch medical imaging AI at scale.
The foundry aims to help vendors develop, validate and then deploy their own AI models without having to start from scratch, said Khan Siddiqui, CEO of Hoppr.
“We provide the platform where health systems, radiology practices and medical device companies can now build their refined models very quickly and deploy them very quickly in their practice or in their product,” he explained.
Hospitals no longer need huge amounts of data or infrastructure to build their own models, because Hoppr and Nvidia pre-train their base models on massive data sets, he pointed out. In the past, providers had to purchase massive datasets containing about 100,000 patient records to train AI models, but pre-trained base models allow hospitals to train models using much smaller datasets, sometimes containing only hundreds of records, Siddiqui said.
The foundry’s goal is to make custom, localized AI development more feasible for vendors, he said.
The focus is on augmenting imaging AI models so providers can integrate specialized tools directly into radiology and diagnostic workflows rather than relying on one-size-fits-all solutions, Siddiqui noted.
David Niewolny, global head of business development at Nvidia, said the AI foundry is a sign of a broader shift, moving healthcare AI from developing isolated models to a full ecosystem of tools that can be deployed directly into clinical workflows.
He said Hoppr solves the “last mile” problem.
“Nvidia provides the tools and the raw performance. Hoppr takes that and, through the use of open models and the adjustments they make, makes it a much more turnkey clinical-grade AI, designed to run inside hospitals,” Niewolny noted.
The partners’ efforts reflect a desire to transform AI in healthcare into something closer to a software development ecosystem than a set of point solutions. They are betting that as core models and deployment platforms mature, vendors will increasingly shift from purchasing AI applications to building and iterating them in-house.
It remains unclear whether this change will actually lead to greater clinical adoption of the tools – or simply add another unnecessary layer of complexity. But it will likely help determine how quickly imaging AI can scale from pilot projects to routine care.
Photo: Peterhowell, Getty Images
