Kalman Filter-Based Novel Methodology to Assess Insulin Pump Accuracy

Area of Science:

• Biomedical Engineering
• Endocrinology
• Metrology

Background:

• Continuous subcutaneous insulin infusion (CSII) systems are vital for type 1 diabetes management and artificial pancreas (AP) development.
• Accurate insulin delivery from CSII systems is critical for glycemic control, yet current assessment methods have limitations, especially at low basal rates.
• Existing gold-standard assessment methods like IEC 60601-2-24 do not sufficiently address the precision of CSII systems at minimal delivery settings.

Purpose of the Study:

• To introduce and validate a novel, accurate, and reactive method for evaluating CSII system performance based on direct flow measurement.
• To assess the delivery precision of an off-the-shelf insulin pump across various basal rates using the new methodology.
• To quantify CSII dose errors, particularly at low basal rates relevant to pediatric use and AP integration.

Main Methods:

• Employed a leading-edge assessment approach combining a direct mass flow meter and a time-stamped microgravimetric bench test.
• Integrated a Bayesian-based mathematical filter (Kalman filter) to enhance measurement accuracy and reactivity.
• Evaluated CSII dose errors using mean absolute relative dispersion (MARD) across different time intervals (15min and 240min) and basal rates (2, 0.5, and 0.1 UI/h).

Main Results:

• The new method demonstrated consistent metrological performance and filtering effectiveness.
• Insulin pump precision varied significantly with flow rate: MARD at 15min was 12.7% (2 UI/h), 20.4% (0.5 UI/h), and 65.0% (0.1 UI/h).
• Longer-term MARD at 240min showed 8.1% (2 UI/h), 18.8% (0.5 UI/h), and 18.4% (0.1 UI/h), with instantaneous flow rate results indicating irregular stroke-based delivery.

Conclusions:

• The validated novel method highlights significant imprecision in current insulin pump delivery, particularly at the lowest basal rates.
• This imprecision is especially concerning for pediatric CSII applications and artificial pancreas systems that rely on minimal insulin delivery.
• The proposed direct flow measurement technique should be adopted for precise comparative analysis of CSII system performance in critical low-dose delivery scenarios.