对基于变压器的模型进行多地平线血糖预测的比较研究
Meryem Altin Karagoz1, Marc D Breton1, Anas El Fathi1
1Center for Diabetes Technology, the University of Virginia, Charlottesville, VA, USA.
ArXiv
|June 4, 2025
概括
变压器模型在1型糖尿病中对血糖 (BG) 预测有希望. 像Crossformer和PatchTST这样的补丁智能变压器,使用长达一周的历史数据实现了卓越的准确性.
科学领域:
- 生物医学工程 生物医学工程
- 人工智能的人工智能
- 内分泌学 在内分泌学.
背景情况:
- 准确的血糖 (BG) 预测对于管理1型糖尿病至关重要.
- 变压器模型为复杂的时间序列预测提供了高级功能.
- 它们对BG预测的应用是一个新兴的研究领域.
研究的目的:
- 为了比较分析变压器模型用于多地平线BG预测.
- 评估不同的嵌入策略 (点智能,补丁智能,系列智能,混合) 用于BG预测.
- 评估输入历史长度 (长达1周) 对预测准确性的影响.
主要方法:
- 使用了公开可用的数据集:DCLP3 (n=112) 和OhioT1DM (n=12).
- 经过训练的变压器网络使用持续血糖监测 (CGM),胰岛素和食数据.
- 评估了短期 (30分钟) 和长期 (1-4小时) BG预测视界的模型.
主要成果:
- 在30分钟的BG预测 (RMSE15.6 mg/dL在OhioT1DM上) 中,交叉变形器 (补丁智能) 卓越.
- 补丁TST (补丁智能) 在1-4小时预测方面表现出卓越的性能 (RMSEs在OhioT1DM上为24.6-46.5 mg/dL).
- 基于补丁的代币化和更长的输入历史记录 (1周) 一般提高了预测准确性.
结论:
- 变压器架构,特别是补丁式方法,显示了准确的BG预测的巨大潜力.
- 这些模型有效地捕捉了糖尿病管理的多变量时间序列数据中的复杂模式.
- 对变压器模型的进一步研究可以增强1型糖尿病的个性化干预措施.
相关概念视频
01:16Hormones Regulating Blood Glucose
Insulin is released by beta cells of the pancreas when blood glucose levels are high. It facilitates glucose absorption and utilization in insulin-dependent cells with insulin receptors on their plasma membranes. Insulin promotes glucose uptake by increasing the number of glucose transport proteins in the cell membrane, allowing glucose to enter the cell. As a result, glucose utilization and ATP production are enhanced.
In addition to accelerating glucose uptake and utilization, insulin has...
In addition to accelerating glucose uptake and utilization, insulin has...
01:26Pharmacokinetic Models: Comparison and Selection Criterion
Physiological and compartmental models are valuable tools used in studying biological systems. These models rely on differential equations to maintain mass balance within the system, ensuring an accurate representation of the dynamic processes at play.
Physiological models take a detailed approach by considering specific molecular processes. They can predict drug distribution, metabolism, and elimination changes, providing a comprehensive understanding of how drugs interact with the body.
Physiological models take a detailed approach by considering specific molecular processes. They can predict drug distribution, metabolism, and elimination changes, providing a comprehensive understanding of how drugs interact with the body.