Harnessing the power of AI to advance knowledge of type 1 diabetes

Credit: Unsplash/CC0Public Domain

Researchers from Texas Children’s Hospital, Children’s Mercy Kansas City (MU), and the University of Missouri (MU), used a new data-driven approach for learning more about type 1 diabetes patients. Type 1 diabetes is responsible for approximately 5-10% of all diabetes diagnoses. The team used health informatics to gather data and then applied artificial intelligence (AI). This helped them better understand the disease.

The team analysed publicly available real-world data from around 16,000 participants who were enrolled in the T1D exchange clinic registry. The team was able identify major differences in the health outcomes of type 1 diabetics who have or don’t have a family history by using a contrast pattern-mining algorithm that was developed at the MU college of Engineering.

Chi-Ren Shyu (director of the MU Institute For Data Science Informatics) led the AI approach used in this study and stated that it is exploratory in nature.

“Here we let it do the work of connecting millions upon millions of dots in data to identify only major contrasting pattern between individuals with type 1 diabetes and those without,” said Shyu. She is the Paul K. Shumaker Professor in MU College of Engineering.

Erin Tallon is a graduate student at the MUIDSI who was the lead author of the study. She said that some of the results from the analysis were not expected.

Tallon said that individuals who had a close relative with type 1 diabetes were more likely be diagnosed with hypertension, diabetes-related nerve diseases, and kidney disease. “We also found an increased likelihood of these conditions in people who had a family history of type 1. People with type 1 diabetes in their immediate families were also more likely to have certain demographic characteristics.

Tallon’s passion for the project started with a personal connection. Her experience as a nurse in an ICU quickly led to her growing interest. Tallon would often see patients with type-1 diabetes with co-existing conditions like high blood pressure and kidney disease. Because type 1 diabetes diagnosis is usually only made when the disease has advanced, she wanted to find better ways of preventing and diagnosing it.

Mark Clements is a pediatric endocrinologist from Children’s Mercy Kansas City. He is also a professor of pediatrics at University of Missouri Kansas City and the corresponding author of the study. In 2019, he was invited to speak at Midwest Bioinformatics Conference hosted at BioNexus Kansas City. Tallon couldn’t attend Clements’ presentation. However, she continued to communicate her ideas for improving understanding of type 1 diabetes to people by phone. He was intrigued. Tallon eventually introduced Clements and Shyu to each other, and a research partnership was established.

Tallon stated that the results of the collaboration show the power and value in using real-world data.

“Type 1 diabetes is not a single disease that looks the same for everybody—it looks different for different people—and we’re working on the cutting-edge to address that issue,” Tallon said. “By analyzing real world data, we can better identify risk factors that may lead to poor outcomes for someone’s health.

Although the results are encouraging, Tallon stated that researchers were limited because they did not have a population-based dataset to work with.

Tallon stated, “It is important that you note here that our findings do not have a limitation that will be addressed in the future using larger, population-based datasets.” “We are looking to build larger patient populations, analyze more data, and use these algorithms for that purpose.

Personalizing medicine

Clements hopes that the approach can be used to help develop personalized treatment options and support people with diabetes.

“In order to get the right treatment to the right patient at the right time, we first need to understand how to identify the patients who are at a higher risk for the disease and its complications—by asking questions such as if there are characteristics early in someone’s life that can help identify an individual with high risk for an outcome years down the road,” Clements said. “All of this information could help us to one day establish a better picture of a person and allow us to create a more personalized approach to both prevention and treatment.

The study was published in Diabetes Care. The study was also helped by Katrina Boles (MU graduate student) and Danlu Liu (MU graduate student).

Type 1 diabetes patients are more likely than ever to be obese.

More information:
Erin M. Tallon and colleagues, Contrast Pattern mining with the T1D Exchange Registry Reveals Complex Phenotypic Factors, Comorbidity Patterns, and Associated with Sporadic Type 1 Diabetes. Diabetes Care (2022). DOI: 10.2337/dc21-2239

Provided by
University of Missouri

Use AI to improve our understanding of type 1 diabetes (2022, 23 March 21)
Retrieved 23 March 2022
from https://medicalxpress.com/news/2022-03-harnessing-power-ai-advance-knowledge.html

This document is subject of copyright. Except for any fair dealing for private study or research, this document is not copyrighted.
Part may not be reproduced without written permission. The information is provided only for informational purposes.

Source: medical xpress.

Leave a Reply

Your email address will not be published.

Back to top button