NIH Funds Project to Speed Diagnosis of Rare Diseases

Researchers are developing PANDA, a disease prediction software

Patricia Inacio PhD avatar

by Patricia Inacio PhD |

Share this article:

Share article via email
Illustration shows a doctor using a desktop computer as he talks to a patient.

The diagnosis of rare diseases such as akylosing spondylitis (AS) is complex and often delayed. A partnership between U.S. researchers aims to use artificial intelligence to develop a software to diagnose rare diseases sooner.

The four-year project, led by researchers at the University of Florida Health and the Perelman School of Medicine at the University of Pennsylvania (UPenn), is funded by a $4.7 milion from the National Institutes of Health (NIH).

Specifically, researchers will leverage the information already collected in patients’ electronic health records to develop a set of machine learning-powered softwares — a method of data analysis — to identify people at risk of five different types of vasculitis (a condition marked by inflammation of the blood vessels) and two different types of spondyloarthritis, including AS and psoriatic arthritis.

Recommended Reading
An illustration shows a person doing stretching exercises.

Group Pilates Superior to At-home Exercise for Patients in Study

“An earlier diagnosis of any of the types of vasculitis and spondyloarhritis we’re working on leads to a much better prognosis and better clinical outcomes,” Peter A. Merkel, MD, chief of rheumatology, professor of medicine and epidemiology at UPenn and one of the project’s co-lead researchers, said in a press release.

“Even if we determine that a patient has just a 10% likelihood of developing one of these diseases, that is a much higher chance of a rare problem, and clinicians can keep that in mind and make better decisions for their patients,” he added.

The other co-lead researchers are Jiang Bian, PhD, professor and chief data scientist for University of Florida Health and Yong Chen, PhD, a professor of biostatistics at UPenn.

The term rare disease is used to describe disorders affecting less than 200,000 people in the U.S. In the clinics, these disorders are dubbed “zebras” because they are unusual and unexpected. Lack of access to data from other clinicians, and unfamiliarity with rare disorders contributes to further delayed diagnosis.

This means that “some patients with rare diseases may go undiagnosed and untreated for years,” said Bian.

The new partnership will use patients’ electronic health records and artificial intelligence to develop a prediction software — called PANDA (short for Predictive Analytics via Networked Distributed Algorithms for multi-system diseases) — that alerts clinicians when a patient is more likely to have a rare disorder.

A software that automatically scans known information could lead to an early diagnosis, according to the press release.

“The increasing availability of real-world data, such as electronic health records collected through routine care, provides a golden opportunity to generate real-world evidence to inform clinical decision-making,” said Bian. “Nevertheless, to leverage these large collections of real-world data, which are often distributed across multiple sites, novel distributed algorithms like PANDA are much needed.”

“This is an exciting step forward, building on our current PDA [Privacy-preserving Distributed Algorithms] framework from clinical evidence generation toward AI-informed interventions in clinical decision-making,” said Chen.

Data from PCORnet

The researchers will use information from the PCORnet (the National Patient-Centered Clinical Research Network), a large clinical research network register, with data from more than 27 million patients nationwide.

Anonymized clinical data (including lab test results, presence of other diseases, treatment history and other common information) will be used during the first year of the project to create the machine learning algorithms. These algorithms are developed initially with a subset of data, followed by training with data form other sources to improve their accuracy, hence the “learning” term.

Once built the researchers will test its predictive power across more than 10 health systems. The prediction software will be made publicly available and can be applied to other conditions.

“Ultimately, we hope to build on the algorithms developed for rare diseases and apply them to other diseases” said Bian.