Adaptive control algorithm for a rapid and slow acting insulin therapy following run-to-run methodology Conference Paper uri icon

abstract

  • An adaptive scheme is developed to automatically adjust the insulin dosing protocol for a person with type 1 diabetes mellitus. The adaptive strategy follows the run-to-run philosophy with a multiple-daily injections (MDI) therapy. Insulin lispro (rapid acting) and insulin glargine (slow acting) are employed in the protocol which uses pre and postprandial glucose concentrations. The key information for the synthesis of the control algorithm is the subject%27s insulin sensitivity that is calculated considering that there is no previous information about the subject%27s response to the insulin protocol. Therefore, the sensitivity information is estimated recursively using on- line data in a time scale of days. After the sensitivity is recalculated, the run-to-run correction scheme is updated, obtaining an adaptive MDI therapy. The advisory algorithm is evaluated in silico by constant random parameters variations and superimposing sinusoidal oscillations on glucose-insulin model parameters to implement intra-individual variability of the glucoregulatory system. For this purpose, the glucose-insulin model developed by Dalla Man et al. (2007) and subcutaneous insulin absorption description by Tarin et al. (2005) were employed. The results show that the algorithm is successful in achieving an euglycemic control despite variable meals (±15 %25 variation in carbohydrate content and timing ±15 min) and time-varying parametric variations in the glucose-insulin model. No severe hypoglycemic (< 50 mg/dL) or hyperglycemic (> 180 mg/dL) events were observed in average for 30 virtual patients. © 2010 AACC.
  • An adaptive scheme is developed to automatically adjust the insulin dosing protocol for a person with type 1 diabetes mellitus. The adaptive strategy follows the run-to-run philosophy with a multiple-daily injections (MDI) therapy. Insulin lispro (rapid acting) and insulin glargine (slow acting) are employed in the protocol which uses pre and postprandial glucose concentrations. The key information for the synthesis of the control algorithm is the subject's insulin sensitivity that is calculated considering that there is no previous information about the subject's response to the insulin protocol. Therefore, the sensitivity information is estimated recursively using on- line data in a time scale of days. After the sensitivity is recalculated, the run-to-run correction scheme is updated, obtaining an adaptive MDI therapy. The advisory algorithm is evaluated in silico by constant random parameters variations and superimposing sinusoidal oscillations on glucose-insulin model parameters to implement intra-individual variability of the glucoregulatory system. For this purpose, the glucose-insulin model developed by Dalla Man et al. (2007) and subcutaneous insulin absorption description by Tarin et al. (2005) were employed. The results show that the algorithm is successful in achieving an euglycemic control despite variable meals (±15 %25 variation in carbohydrate content and timing ±15 min) and time-varying parametric variations in the glucose-insulin model. No severe hypoglycemic (< 50 mg/dL) or hyperglycemic (> 180 mg/dL) events were observed in average for 30 virtual patients. © 2010 AACC.

publication date

  • 2010-01-01