Scholarlink

Continuous blood glucose monitoring offers a wealth of data for type 2 diabetes management. In this pilot study, we explored the potential of Long Short-Term Memory (LSTM) models to predict blood sugar levels. We monitored 10 diabetic and 10 non-diabetic individuals for 14 days, using their high-frequency glucose data to train the model. The model successfully tracked real glycemic fluctuations in diabetic individuals, anticipating trends 5 h ahead with 12-h data. Predicted values aligned with carbohydrate intake, physical assessments, observed peaks, lifestyles, and daily carbohydrate consumption. The model faced challenges with non-diabetic data due to its inherently lower variability compared to diabetic data with its characteristic spikes. While the model captured the overall peak trend (1:15–1:30 pm) for the entire group, a larger study population is needed to refine and generalize the model for broader application.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting Blood Glucose Levels in Type II Diabetes: A Pilot Study Using Continuous Monitoring and Long Short-Term Memory Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • M. H. V. Mourão,
  • L. dos Santos,
  • M. M. Amaral

摘要

Continuous blood glucose monitoring offers a wealth of data for type 2 diabetes management. In this pilot study, we explored the potential of Long Short-Term Memory (LSTM) models to predict blood sugar levels. We monitored 10 diabetic and 10 non-diabetic individuals for 14 days, using their high-frequency glucose data to train the model. The model successfully tracked real glycemic fluctuations in diabetic individuals, anticipating trends 5 h ahead with 12-h data. Predicted values aligned with carbohydrate intake, physical assessments, observed peaks, lifestyles, and daily carbohydrate consumption. The model faced challenges with non-diabetic data due to its inherently lower variability compared to diabetic data with its characteristic spikes. While the model captured the overall peak trend (1:15–1:30 pm) for the entire group, a larger study population is needed to refine and generalize the model for broader application.