Fig 1.
(Right) Users can specify ICR and insulin basal rate pump settings. (Left) Simulation Results displayed include glucose readings over a 10-minute interval increments and display meal amounts, basal rates, and insulin boluses.
Fig 2.
Algorithm membership function.
(Top) Input Membership Function. The algorithm classifies glucose input into 4 sets: low, medium, high, and ex_high. The x-axis plots blood glucose values and the y-axis plots the range of truth value from 0 to 1. (Bottom) Output Membership Function involved in the center of gravity defuzzification process to produce basal-rate insulin changes.
Fig 3.
Algorithm insulin basal rate suggestions.
Upon analyzing a 48-hour simulation run, the Fuzzy-Logic algorithm returns suggested changes to basal rates. Values displayed in green prompt the user to a necessary increase in the pump parameter, red values suggest decreasing a basal rate, and black values suggest no change.
Fig 4.
The suggested ICR values in green indicate an increase from the current ICR and suggested ICR values in red indicate a decrease from the current ICR.
Fig 5.
Comparison of Open-Loop vs. Fuzzy-Logic Learning algorithms.
(Top) 31-day trial using the Open-Loop Algorithm (bottom) Fuzzy-Logic Learning Algorithm trial results. Red Lines represent the 25% and 75% IQR median range of blood glucose levels across all 15 patients, the blue lines represent insulin basal injections, and the x-axis plots clinical trial simulation time in hours.
Table 1.
CSII Open-Loop vs CSII Fuzzy-Logic Learning Algorithms comparison for the overall simulation study period of 31 days across the 15 patients.
Table 2.
Training session results for 13 nurses spanning training stages 2–4.