Categories | Inventors
Technology Overview
OHSU # 2705 — Glucose prediction algorithm using long-short term memory recurrent neural network
Summary
The OHSU glucose prediction algorithm is a data-driven glucose prediction model trained on a big dataset to predict glucose concentration within a short term (30 minute) period for patients with type 1 diabetes.
Technology Overview
Patients with type 1 diabetes do not produce their own insulin and rely on continuous glucose monitoring (CGM) systems and insulin pumps to help manage glucose levels. Accurate glucose prediction algorithms are critical components of CGM systems to help people proactively avoid adverse hyper- or hypo-glycemic events.
The OHSU glucose prediction algorithm is a data-driven glucose prediction model trained on a big dataset to predict glucose concentration within a short term (30 minute) period. The model is composed of a recurrent neural network with long-short-term memory units and a patient specific smoothing error correction step. The OHSU algorithm can be integrated into continuous glucose management-based decision support tools to alert insulin pump Type 1 diabetes users of glycemic changes. In addition to use in CGM systems, the OHSU algorithm can also be integrated into artificial pancreas algorithms.
Publication
Mosquera-Lopez C, Jacobs PG. Incorporating Glucose Variability into Glucose Forecasting Accuracy Assessment Using the New Glucose Variability Impact Index and the Prediction Consistency Index: An LSTM Case Example. J Diabetes Sci Technol. 2022 Jan;16(1):7-18. Link
Licensing Opportunities
Available for non-exclusive and exclusive licensing.
Inventors:
Patents
| Filed | United States |
Categories: