Accuracy of Noninvasive Glucose Sensing Based on Near-Infrared Spectroscopy.
jianjun liu,Rong Liu,Kexin Xu
DOI: https://doi.org/10.1366/14-07728
2015-01-01
Applied Spectroscopy
Abstract:The noninvasive sensing of the blood glucose concentration is usually based on optical, electrical, or acoustical signals induced by blood glucose; these signals are extremely weak and subject to fluctuations caused by the variation in the body or surroundings. Therefore, it is challenging to detect blood glucose noninvasively with high accuracy, and no successful accurate and noninvasive clinical application has been reported. We found that there are two key measurement issues to be addressed: systematic errors, such as the errors induced by the drifts of devices or by variations in body temperature, among others, are too large to guarantee the trueness of measurement at present; and random disturbances in repeated tests, such as disturbances associated with variations in the human-machine interface, pulses, and the thermal noise of the devices, cause larger repeated measurement errors and compromise precision. Recent novel reference measurements based on differential near-infrared (NIR) spectroscopy are considered promising for solving the systematic error issue by establishing matched references, collected at another detection site or at another time, and subsequently differencing to remove the common systematic errors. However, differencing weakens the signal of interest itself and enlarges the effects of the second issue, random disturbances affecting the precision. It is understood that only reference measurements that can meet the precision requirement will be promising for future applications. Therefore, this study quantitatively evaluates the precision of the main differential NIR spectroscopy measurements considering similar conditions and minimized random disturbances. The precision of the measurements under these conditions should represent their optimal precision levels. After the evaluation, noninvasive glucose-sensing methods that hold promise for future clinical application are proposed. Finally, the evaluation criteria could be a reference for the noninvasive detection of other physiological components.