ML Tool Speeds Deployment of Health Predictor
Large health datasets are being used to develop predictive risk models for individual and population groups. The latest example is a partnership between a predictive analytics vendor and a machine-learning platform specialist to deploy a new health predictor.
Brooklyn-based Yhat, developer of a machine-learning deployment platform called ScienceOps, said this week that analytics vendor Lumiata is using its platform to launch its AI-powered health prediction tool. Yhat said its platform helps overcome incompatibilities between AI algorithms and emerging digital applications.
ScienceOps is positioned as providing the technical infrastructure “to transform statistical code on an analyst’s laptop into a product you and I can interact with,” explains Austin Ogilvie, Yhat’s CEO and co-founder. The goal is to help launch more AI-based applications, Oglivie added.
Lumiata, San Mateo, Calif., is among the growing number of predictive analytics developers targeting medical AI applications to manage risk and prioritize healthcare resources. The company used the Yhat platform to incorporate its proprietary health risk algorithms into a predictive tool called the Risk Matrix. The tool is designed to deliver “personalized, time-based predictions of an individual’s future health state based on associated clinical conditions or diagnoses,” the partners said.
The Yhat deployment platform zeroes in on a recurring problem for data scientists: open source statistical tools are often incompatible with frameworks and languages used to build applications. Yhat claims its ScienceOps platform allows data scientists to leverage algorithms written in R or Python directly within mobile or web apps.
Lumiata said its AI-based predictive models are based on huge and complex datasets that must be continuously refined to deliver relevant risk predictions. Those predictions are delivered via an API. Along with fine-tuning its models, the ScienceOps platform runs on-premise to ensure compliance with medical privacy regulations such as HIPAA (Health Insurance Portability and Accountability Act).
Lumiata’s touts its medical AI tools as meshing multi-sourced health data and then applying graph models, natural language processing and deep learning to deliver medically relevant analytics. The company’s medical graph tool combines data derived from medical research and clinical practice with multi-sourced health data to analyze the relationships among them. The goal is delivering “hyper-personalized” business and clinical insights.
Meanwhile, Yhat’s software platform deploys predictive algorithms as REST APIs while eliminating the engineering tasks associated with production environments such as testing, scaling and security.
The collaboration illustrates how emerging medical analytics and healthcare IT startups that have been attracting large private equity and corporate venture capital investments are beginning to roll out new tools targeting specific use cases. For example, Health Catalyst, based in Salt Lake City, raised $70 million during the first quarter of this year to advance its technology platform used to organize and link health-related data stored on different platforms.