(625h) Development of a Model-Based Noninvasive Continuous-Time Glucose Monitoring Device for Non-Insulin Dependent People

Continuous-time glucose monitoring (CGM) effectively
improves glucose control by providing frequently sampled information that
allows the user to associate changes in their glucose levels with changes in
their behavior. For example, a user is able to see immediately how the size of
a meal affects how high their glucose level becomes as well as the duration of
elevated glucose levels. Currently,
the most widely used and effective CGM devices rely on a sensor that is
inserted invasively under the skin. Sensors cost from $35 to $60 and last 3
days to a week. Due to invasiveness and cost, the primary users of current CGM
devices are insulin dependent people (type 1 and some type 2 diabetics). Thus,
the primary goal of this research is the development of a non-invasive CGM
device that would be used by health conscious non-insulin dependent people
(including non-diabetics) that would help reduce obesity, the onset of type 2 diabetes, as well as the progression of type 2
diabetes. 

To accomplish this objective, this work is focusing on the
development of a device that: 1. has a simple or no reliance for on food entry;
2. has a relatively short calibration period; 3. requires few to no lancet
measurements for calibration and; 4. has an accuracy that is comparable to
lancet meters. The approach of this research is to use a novel modeling method
technique to infer glucose concentration using non-invasive input measurements
from variables representing food, activity, circadian rhythm, and stress
variations. The main component of this system will be a the
BodyMedia® armband that will automatically collect
the activity and stress data. The food information will be entered manually by
the user via the time stamp button on the armband. The model and model
development algorithm will reside in the armband and will use a Wiener approach
based on the work of Rollins et al. (2010). Data from a lancet glucose  meter will be
entered automatically or manually  and
will be used to develop a subject specific model for the person wearing the
device. After the model is completely developed, lancet measurements will no
longer be needed for calibration. An interface device will be connected to the
armband to display the glucose concentration in every five minutes.

Using about 20 test subjects with 4 weeks of data collection
each, results have been obtained to support the modeling viability necessary to
build an armband monitoring device. Before using these data sets the food
quantities were converted to food indices to mimic time stamping and a lancet
sampling rate of only four values per day. Several improvements to the Wiener
modeling method developed by Rollins et al. were made to address the infrequent
nature of the data (i.e., four glucose measurements/day), frequent start- ups
(i.e., taking the armband on and off and restarting it frequently), and the
lack of steady state data. The accomplishments of this work include the ability
to develop subject specific models under these restrictions. Results are
presented for models developed after three days, 2 weeks and four 4 weeks that
support an initial calibration period of three days with accuracy improving
over time and no need for lancet measurements after 3 to 4 weeks. Thus, since
the model does not drift, this device would not appear to need glucose
measurements for calibration once it is fully calibrated to the user wearing
the device.

Rollins, D. K., N. Bhandari,
J. Kleinedler, K. Kotz, A. Strohbehn, L. Boland, M. Murphy, D. Andre, N. Vyas,
G. Welk and W. Franke, "Free-living inferential modeling of blood glucose
level using only noninvasive inputs," Journal of Process Control 20 95-107 (2010).