Pattern Classification (2nd Edition)
Pattern Classification (2nd Edition)
| Hi-index | 0.00 |
A novel method has been designed to estimate blood glucose variations noninvasively. The concept of this method is based on the glucose metabolic process in the human body. The data acquisition phase was implemented by measuring metabolic parameters including heat dissipation by conduction at the fingertip, percentage oxygen content of expired air and minute volume of expired air. A data analysis phase was performed to convert the measured parameters to variations of features, which would become input of the calibrated classification model for estimation of blood glucose variations. The classification model can be a 2-class to 5-class system showing different resolutions for the extents of blood glucose variations. In the experimental trial, a total of 190 data points were obtained from 31 normal and 159 type 2 diabetic subjects. The classification accuracy for a 3-class system was 84.26% using a linear discriminant classifier. Multiple regression analysis was also performed to compare the noninvasive method with variations of glucometer readings. The correlation coefficient was 0.88. Preliminary results show that this method has the potential to be used as blood glucose variation monitors and lifestyle educating devices for normal, pre-diabetic and type 2 diabetic persons.