How NIH Is Translating 70 Years of Well being Information To Communicate the Identical Language

12 Petabytes and a Converter Field

On the middle of the hassle is BioData Catalyst, a cloud-based ecosystem developed by NHLBI in collaboration with NLM and ODSS. 

“[NIH has] over 12 petabytes of knowledge that’s all multimodal — the whole lot from genomics, medical imaging, sleep, sensor information, all totally different modalities,” mentioned Sweta Ladwa, chief of the Scientific Options Supply Department on the Information Technology and Purposes Heart (ITAC) at NHLBI. That information spans long-running research such because the Trans-Omics for Precision Medication (TOPMed) program, which tracks roughly 180,000 people.

However entry alone doesn’t make information AI-ready. The harder downside is interoperability — guaranteeing a cardiovascular variable from, say, a Nineties Framingham Coronary heart Examine cohort means the identical factor as one from a current pulmonary fibrosis research. NHLBI has constructed a linked information modeling language (LinkML) pipeline to unravel this — what Ladwa referred to as a “converter field strategy, the place you’ll be able to plug within the information from the supply and IT will put IT in that [format] to your evaluation.” The pipeline maps information throughout a number of requirements, together with LOINC (Logical Commentary Identifiers Names and Codes), FHIR (Quick Healthcare Interoperability Sources) and HPO (Human Phenotype Ontology).

NHLBI pairs the automated mapping with medical validation. “We’ve pulmonologists who we’re working with to essentially clinically decide the ideas,” Ladwa mentioned, “as a result of we wish to guarantee that this fancy hypertensive remedy is similar as this different one. They’re all in the identical class, they usually’re all the identical ontological idea.” She added that the AI-assisted mapping makes use of “publicly accessible metadata” slightly than affected person information.

READ MORE: Healthcare interoperability improves care and affected person experiences.

From Analysis Requirements to the Digital Medical Report

Whereas NHLBI focuses on making present analysis information interoperable, ODSS is engaged on a complementary downside: getting research-grade requirements into the medical techniques the place new information is generated day by day.

Susan Gregurick, NIH’s affiliate director for information science and director of ODSS, described a push to map NIH analysis requirements into the USA Core Information for Interoperability (USCDI) — the interoperability customary that digital medical file techniques use for accreditation. As a part of this course of, Gregurick cited work that started in oncology and is increasing to different illness areas, together with a cardiovascular partnership with NHLBI.

The implication: When cardiovascular phenotypes seem in a affected person encounter — even exterior a proper research — EMR techniques can seize them in a format researchers can use. 

“The affect for that kind of cross-agency collaboration is actually big,” Gregurick mentioned. “I feel that that’s nearly other than AI, however IT’s going to be one thing that drives AI sooner or later.”

Click on the banner under to join HealthTech’s weekly publication.