Researchers have secured a $1.2 million NIH grant through 2029 to deploy artificial intelligence models for managing Type 2 Diabetes, using Restless Multi-Armed Bandit frameworks to optimize patient care.
Grant Funds Artificial Intelligence for Diabetes Care
Decentralized health systems often share responsibility for the same population in an area, and sometimes even compete for the same medications or funding. Public health systems constantly battle the logistical hurdle of distributing limited intervention funding and resources where they will do the most good. A collaborative team of computer scientists and medical professionals aims to tackle this issue by deploying specialized artificial intelligence models designed to determine who should receive care first. Patient needs, varying disease trajectories, and social determinants such as healthcare accessibility make these choices even more intricate.
The National Institutes of Health awarded the group an R01 grant. The funding totals $1.2 million and extends through 2029 to support the development and implementation of these predictive frameworks.
Our goal is to address the immense challenge of allocating limited health intervention resources in optimal ways.
Kai Wang, assistant professor in the School of Computational Science and Engineering, via Gatech
Wang leads the Georgia Tech side of the project alongside fellow School of Computational Science and Engineering Professor B. Aditya Prakash. Multi-agent models are aimed at decentralized health systems to make fairer decisions by examining how to coordinate resources and decisions between stakeholders.
Researchers Target Complexities of Type 2 Diabetes
Diabetes is a growing health concern in developing countries, causing considerable mortality rates.
Diabetes is a great example of a complicated, multi-faceted problem. It is a natural playground for the kind of high-dimensional machine learning algorithms we want to build and deploy.
B. Aditya Prakash, professor in the School of CSE, via Gatech
As a chronic disease, diabetes also creates opportunities for sequential decision-making. Clinicians must address an ongoing sequence of choices over an extended period instead of selecting a single treatment option.
Institutions Collaborate on Clinical Validation
The team’s framework applies Restless Multi-Armed Bandit models across three different thrusts to help providers and healthcare systems treat diabetes patients.
We lean on clinical experts to provide the data and expertise needed to guide our methods and make sense of results. Ultimately, these approaches need to be clinically validated if we want to make real-world progress.
B. Aditya Prakash, professor in the School of CSE, via Gatech
A study from two Chinese hospitals suggests that machine learning could help identify patients with biopsy-confirmed non-alcoholic fatty liver disease who are more likely to develop Type 2 diabetes. Researchers evaluated seven machine learning approaches, including logistic regression, random forest, an artificial neural network, and XGBoost, using five routinely available clinical measures selected via LASSO regression—waist circumference, fasting plasma glucose, triglycerides, family history of diabetes, and neutrophil-to-lymphocyte ratio. XGBoost proved superior across the external cohort and became the chosen final model, attaining an area under the receiver operating characteristic curve score of 0.99 during training alongside 0.89 in external validation, while the artificial neural network yielded the top results in internal testing.