Use of Change-Point Analysis Tools for Evaluating Performance in Artificial Pancreas Systems

In this work, we evaluate the potentials and limitations of Change-Point Analysis (CPA) tools for evaluating performance in Artificial Pancreas (AP) systems. In particular, we use Mood and Mann-Whitney's non-parametric tests for detecting changes in scale and location in glucose data distributions. These tools are applied to the results of the clinical trials that took place in June 2017 at the Hospital Italiano de Buenos Aires (HIBA) where the first in vivo tests of the Automatic Regulation of Glucose (ARG) control algorithm were carried out.