Development of urine glucose level monitor for home healthcare using near infrared spectroscopy

We have been developing a urine glucose level monitoring technique using near infrared spectroscopy in conjunction with the chemometric method aiming for the use of home health care. In this study, validity of the calibration models were assessed using urine samples from nine healthy male subjects. For the individual measurement, reasonable accuracy was obtained in predicting urine glucose level for each subject (mean value of standard error of prediction; 25.8 mg/dl, S.D.; 10.8 mg/dl), however, the accuracy decreased (around 60 mg/dl) when the calibration models were generated using data from multiple subjects. One of the causative factors seemed to be urine urea concentration level which may vary within very wide range (0-2000 mg/dl) under physiological condition, and therefore, it was suggested that any idea, e.g., classification of data by urea level, would be required for improving the accuracy.