首页> 外文会议>Biological and artificial computation : From neurosciene to technology >Neural Network-Based Insulin Infusion Control for an Insulin-Pump, Using Discontinuous Blood glucose Measurements
【24h】

Neural Network-Based Insulin Infusion Control for an Insulin-Pump, Using Discontinuous Blood glucose Measurements

机译:使用不连续血糖测量的基于神经网络的胰岛素泵胰岛素输注控制

获取原文
获取原文并翻译 | 示例

摘要

This paper presents a neural network-based system for the computation of appropriate hourly insulin doses for insulin-pumps adjustments, using short historical discontinuous measurements of blood glucose levels. Our database consists of 25000 records of blood glucose measurements and corresponding insulin doses levels adjustments. Ten discontinuous measurements per day wer carried out on 747 patients under pump treatment. In order to predict the next-time insulin-dose, one neural network for each period have been trained. So, each one of the ten neural networks specialized to a specific period of the day. The efficient data concept is introduced. Training with efficient learning data allowed to achieve very good generalization on both efficient and nonefficient data. A computer program based on the trained neural networks is under clinical validation test. A neural network-based injection-pump is also beeing prototyped.
机译:本文介绍了一种基于神经网络的系统,该系统使用短期历史上不连续的血糖水平测量来计算适当的每小时胰岛素剂量以进行胰岛素泵调整。我们的数据库包含25000条血糖测量记录和相应的胰岛素剂量水平调整。每天对747名接受泵治疗的患者进行10次不连续测量。为了预测下一次的胰岛素剂量,已经训练了每个时期的一个神经网络。因此,十个神经网络中的每个神经网络都专门针对一天中的特定时间段。介绍了有效的数据概念。通过使用有效的学习数据进行培训,可以对有效和无效数据进行很好的概括。基于受过训练的神经网络的计算机程序正在临床验证测试中。基于神经网络的注射泵也正在被原型化。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号