Near-infrared noninvasive blood glucose prediction without using multivariate analyses: introduction of imaginary spectra due to scattering change in the skin.

A noninvasive measurement method is proposed and examined to continuously predict blood glucose contents using near-infrared diffuse reflection difference spectra measured at the skin tissue without using multivariate analyses. Using the modified Beer's law, the difference spectra are assumed to be synthesized from four major components in the human skin (water, protein, glucose, and fat) and a scattering equivalent component called baseline. As a result, one of the origins of the errors in blood glucose prediction using near-infrared is found to be the similarity of the shapes of the absorption spectrum between glucose and baseline. After separating the glucose contributions from the difference spectra at the characteristic wavelengths of baseline and fat, an imaginary component combining baseline and fat is introduced by considering that both the change in the fat contribution and the generation of baseline originate from the change in scattering in the skin. The imaginary component enables us to reduce the errors in blood glucose prediction. In contrast to the methods using multivariate analyses, the calculation process of the blood glucose contents from the measured reflection spectra is clear in this method, thus, it is easy to estimate the origins of the changes and contributions of the components in the measured difference spectra. The proposed method may become a useful tool for realization of noninvasive blood glucose prediction using near-infrared spectroscopy.

Preliminary evaluation of optical glucose sensing in red cell concentrations using near-infrared diffuse-reflectance spectroscopy.

Bacterial contamination of blood products is one of the most frequent infectious complications of transfusion. Since glucose levels in blood supplies decrease as bacteria proliferate, it should be possible to detect the presence of bacterial contamination by measuring the glucose concentrations in the blood components. Hence this study is aimed to serve as a preliminary study for the nondestructive measurement of glucose level in transfusion blood. The glucose concentrations in red blood cell (RBC) samples were predicted using near-infrared diffuse-reflectance spectroscopy in the 1350 to 1850 nm wavelength region. Furthermore, the effects of donor, hematocrit level, and temperature variations among the RBC samples were observed. Results showed that the prediction performance of a dataset which contained samples that differed in all three parameters had a standard error of 29.3 mg/dL. Multiplicative scatter correction (MSC) preprocessing method was also found to be effective in minimizing the variations in scattering patterns created by various sample properties. The results suggest that the diffuse-reflectance spectroscopy may provide another avenue for the detection of bacterial contamination in red cell concentrations (RCC) products.

Multi-objective genetic algorithm-based sample selection for partial least squares model building with applications to near-infrared spectroscopic data.

In this study, multi-objective genetic algorithms (GAs) are introduced to partial least squares (PLS) model building. This method aims to improve the performance and robustness of the PLS model by removing samples with systematic errors, including outliers, from the original data. Multi-objective GA optimizes the combination of these samples to be removed. Training and validation sets were used to reduce the undesirable effects of over-fitting on the training set by multi-objective GA. The reduction of the over-fitting leads to accurate and robust PLS models. To clearly visualize the factors of the systematic errors, an index defined with the original PLS model and a specific Pareto-optimal solution is also introduced. This method is applied to three kinds of near-infrared (NIR) spectra to build PLS models. The results demonstrate that multi-objective GA significantly improves the performance of the PLS models. They also show that the sample selection by multi-objective GA enhances the ability of the PLS models to detect samples with systematic errors.

New methodology to obtain a calibration model for noninvasive near-infrared blood glucose monitoring.

This paper reports new methodology to obtain a calibration model for noninvasive blood glucose monitoring using diffuse reflectance near-infrared (NIR) spectroscopy. Conventional studies of noninvasive blood glucose monitoring with NIR spectroscopy use a calibration model developed by in vivo experimental data sets. In order to create a calibration model, we have used a numerical simulation of light propagation in skin tissue to obtain simulated NIR diffuse reflectance spectra. The numerical simulation method enables us to design parameters affecting the prediction of blood glucose levels and their variation ranges for a data set to create a calibration model using multivariate analysis without any in vivo experiments in advance. By designing the parameters and their variation ranges appropriately, we can prevent a calibration model from chance temporal correlations that are often observed in conventional studies using NIR spectroscopy. The calibration model (regression coefficient vector) obtained by the numerical simulation has a characteristic positive peak at the wavelength around 1600 nm. This characteristic feature of the regression coefficient vector is very similar to those obtained by our previous in vitro and in vivo experimental studies. This positive peak at around 1600 nm also corresponds to the characteristic absorption band of glucose. The present study has reinforced that the characteristic absorbance of glucose at around 1600 nm is useful to predict the blood glucose level by diffuse reflectance NIR spectroscopy. We have validated this new calibration methodology using in vivo experiments. As a result, we obtained a coefficient of determination, r2, of 0.87 and a standard error of prediction (SEP) of 12.3 mg/dL between the predicted blood glucose levels and the reference blood glucose levels for all the experiments we have conducted. These results of in vivo experiments indicate that if the parameters and their vibration ranges are appropriately taken into account in a numerical simulation, the new calibration methodology provides us with a very good calibration model that can predict blood glucose levels with small errors without conducting any experiments in advance to create a calibration model for each individual patient. This new calibration methodology using numerical simulation has promising potential for NIR spectroscopy, especially for noninvasive blood glucose monitoring.

Moving window cross validation: a new cross validation method for the selection of a rational number of components in a partial least squares calibration model.

A new cross validation method called moving window cross validation (MWCV) is proposed in this study, as a novel method for selecting the rational number of components for building an efficient calibration model in analytical chemistry. This method works with an innovative pattern to split a validation set by a number of given windows that move synchronously along proper subsets of all the samples. Calculations for the mean value of all mean squares error in cross validations (MSECVs) for all splitting forms are made for different numbers of components, and then the optimal number of components for the model can be selected. Performance of MWCV is compared with that of two cross validation methods, leave-one-out cross validation (LOOCV) and Monte Carlo cross validation (MCCV), for partial least squares (PLS) models developed on one simulated data set and two real near-infrared (NIR) spectral data sets. The results reveal that MWCV can avoid a tendency to over-fit the data. Selection of the optimal number of components can be easily made by MWCV because it yields a global minimum in root MSECV at the optimal number of components. Changes in the window size and window number of MWCV do not greatly influence the selection of the number of components. MWCV is demonstrated to be an effective, simple and accurate cross validation method.

Noninvasive near-infrared blood glucose monitoring using a calibration model built by a numerical simulation method: Trial application to patients in an intensive care unit.

We have applied a new methodology for noninvasive continuous blood glucose monitoring, proposed in our previous paper, to patients in ICU (intensive care unit), where strict controls of blood glucose levels are required. The new methodology can build calibration models essentially from numerical simulation, while the conventional methodology requires pre-experiments such as sugar tolerance tests, which are impossible to perform on ICU patients in most cases. The in vivo experiments in this study consisted of two stages, the first stage conducted on healthy subjects as preliminary experiments, and the second stage on ICU patients. The prediction performance of the first stage was obtained as a correlation coefficient (r) of 0.71 and standard error of prediction (SEP) of 28.7 mg/dL. Of the 323 total data, 71.5% were in the A zone, 28.5% were in the B zone, and none were in the C, D, and E zones for the Clarke error-grid analysis. The prediction performance of the second stage was obtained as an r of 0.97 and SEP of 27.2 mg/dL. Of the 304 total data, 80.3% were in the A zone, 19.7% were in the B zone, and none were in the C, D, and E zones. These prediction results suggest that the new methodology has the potential to realize a noninvasive blood glucose monitoring system using near-infrared spectroscopy (NIRS) in ICUs. Although the total performance of the present monitoring system has not yet reached a satisfactory level as a stand-alone system, it can be developed as a complementary system to the conventional one used in ICUs for routine blood glucose management, which checks the blood glucose levels of patients every few hours.

Improvement of the partial least squares model performance for oral glucose intake experiments by inside mean centering and inside multiplicative signal correction.

Near infrared (NIR) spectroscopy has become a promising technique for the in vivo monitoring of glucose. Several capillary-rich locations in the body, such as the tongue, forearm, and finger, have been used to collect the in vivo spectra of blood glucose. For such an in vivo determination of blood glucose, collected NIR spectra often show some dependence on the measurement conditions and human body features at the location on which a probe touches. If NIR spectra collected for different oral glucose intake experiments, in which the skin of different patients and the measurement conditions may be quite different, are directly used, partial least squares (PLS) models built by using them would often show a large prediction error because of the differences in the skin of patients and the measurement conditions. In the present study, the NIR spectra in the range of 1300-1900 nm were measured by conveniently touching an optical fiber probe on the forearm skin with a system that was developed for in vivo measurements in our previous work. The spectra were calibrated to resolve the problem derived from the difference of patient skin and the measurement conditions by two proposed methods, inside mean centering and inside multiplicative signal correction (MSC). These two methods are different from the normal mean centering and normal multiplicative signal correction (MSC) that are usually performed to spectra in the calibration set, while inside mean centering and inside MSC are performed to the spectra in every oral glucose intake experiment. With this procedure, spectral variations resulted from the measurement conditions, and human body features will be reduced significantly. More than 3000 NIR spectra were collected during 68 oral glucose intake experiments, and calibrated. The development of PLS calibration models using the spectra show that the prediction errors can be greatly reduced. This is a potential chemometric technique with simplicity, rapidity and efficiency in the pretreatment of NIR spectra collected during oral glucose intake experiments.

Removal of interference signals due to water from in vivo near-infrared (NIR) spectra of blood glucose by region orthogonal signal correction (ROSC).

A novel chemometric method, region orthogonal signal correction (ROSC), is proposed and applied to pretreat near-infrared (NIR) spectra of blood glucose measured in vivo. Water is the most serious interference component in such kinds of noninvasive measurements, because it shows very high absorbance in the spectra. In the present study, the spectra of blood glucose in the range of 1212 - 1889 nm are used, in which the absorption of water around 1440 nm is very high. ROSC aims at removing the interference signal due to water from the spectra by selecting a set of spectra with a special region of 1404 - 1454 nm that mainly contain information about the variation of the interference component, water, and calculating the orthogonal components to the concentrations of glucose that will be removed. The difference between ROSC and orthogonal signal correction (OSC) is that ROSC uses a special region of spectra for the estimation of scores and loading weights of orthogonal components to pretreat the spectra in other regions, while OSC only uses one fixed region of spectra to calculate loadings, scores and weights of OSC components and removes the OSC components in the same region. A clear advantage of ROSC is that it is more interpretable than OSC, because one can select a spectral region to remove the variation of a special component such as water. Another chemometric method, moving window partial least squares (MWPLSR), is also used to select informative regions of glucose from the NIR spectra of blood glucose measured in vivo, leading to improved PLS models. Results of the application of ROSC demonstrate that ROSC-pretreated spectra including the whole spectral region of 1212 - 1889 nm or an informative region of 1600- 1730 nm selected by MWPLSR provide very good performance of the PLS models. Especially, the later region yields a model with RMSECV of 15.8911 mg/dL for four PLS components. ROSC is a potential chemometric technique in the pretreatment of various spectra.

In vivo noninvasive measurement of blood glucose by near-infrared diffuse-reflectance spectroscopy.

This paper reports in situ noninvasive blood glucose monitoring by use of near-infrared (NIR) diffuse-reflectance spectroscopy. The NIR spectra of the human forearm were measured in vivo by using a pair of source and detector optical fibers separated by a distance of 0.65 mm on the skin surface. This optical geometry enables the selective measurement of dermis tissue spectra due to the skin's optical properties and reduces the interference noise arising from the stratum corneum. Oral glucose intake experiments were performed with six subjects (including a single subject with type I diabetes) whose NIR skin spectra were measured at the forearm. Partial least-squares regression (PLSR) analysis was carried out and calibration equations were obtained with each subject individually. Without exception among the six subjects, the regression coefficient vectors of their calibration models were similar to each other and had a positive peak at around 1600 nm, corresponding to the characteristic absorption peak of glucose. This result indicates that there is every possibility of glucose detection in skin tissue using our measurement system. We also found that there was a good correlation between the optically predicted values and the directly measured values of blood samples with individual subjects. The potential of noninvasive blood glucose monitoring using our methodology was demonstrated by the present study.