Noninvasive Measurement of Glucose in Artificial Plasma with Near-Infrared and Raman Spectroscopy Jintao Xue,a Han Chen,a Dongmei Xiong,a Guo Huang,b Hong Ai,b Yan Liang,a Xinyu Yan,a Yuan Gan,a Cong Chen,a Ruobing Chao,a Liming Yea,* a b

West China School of Pharmacy, Sichuan University, Chengdu 610041, People’s Republic of China Bruker Instruments Ltd, Beijing 100081, People’s Republic of China

The goal of this research was to develop a method for noninvasive blood glucose assay. Near-infrared (NIR) spectroscopy and Raman spectroscopy, two more promising techniques compared to other methods, were investigated in two kinds of artificial plasma (AP). Calibration models were generated by performing partial least squares (PLS) regression and optimized individually by considering spectral range, spectral pretreatment methods, and number of model factors. The two spectroscopic models were validated for the determination of glucose, and the results show that the two spectroscopic models established are robust, accurate, and repeatable. Compared to Raman spectroscopy, the performance of NIR spectroscopy was much better, with lower root mean square errors of cross-validation (RMSECV) of 0.128 and 0.094 mg/ml, lower root mean square errors of validation (RMSEP) of 0.061 and 0.046 mg/ml, higher correlation coefficients (R) of 99.15% and 99.55%, and higher residual predictive deviations (RPD) of 10.8 and 15.0 for artificial plasma I and II, respectively. Index Headings: Near-infrared spectroscopy; NIR; Raman spectroscopy; Blood glucose noninvasive measurement; artificial plasma; partial least squares; PLS.

INTRODUCTION Diabetes is now taking its place as one of the main threats to human health, which has been declared a global epidemic by the World Health Organization (WHO). The past two decades have seen an explosive increase in the number of people diagnosed with diabetes worldwide. According to WHO, there were over 171 million diabetics worldwide in 2000 (2.8% of the population). It is projected that the number of patients will increase to 366 million by 2030.1–4 Diabetes mellitus is a chronic and incurable disease.3 There are two main forms of diabetes. Type 1 diabetes is due primarily to autoimmune-mediated destruction of pancreatic b-cell islets, resulting in absolute insulin deficiency. People with type 1 diabetes must take exogenous insulin for survival to prevent the development of ketoacidosis. Type 2 diabetes mellitus is far more common, which accounts for over 90% of cases globally. Insulin resistance and/or abnormal insulin secretion, either of which may predominate, characterize type 2 diabetes mellitus. People with type 2 diabetes are not dependent on exogenous insulin, but may require it for control of blood glucose levels if this is Received 9 August 2013; accepted 27 November 2013. * Author to whom correspondence should be sent. E-mail: [email protected]. DOI: 10.1366/13-07250

428

Volume 68, Number 4, 2014

not achieved with diet alone or with oral hypoglycemic agents.1,5,6 Therefore, inadequate secretion of insulin by the diabetic pancreas results in poor maintenance of normoglycemia (defined as the normal condition of blood glucose concentrations within the 70–100 mg/dL range) with elevated blood glucose concentrations, sometimes upward of 300 mg/dL.7 Unregulated blood glucose concentrations can cause many medical complications including diabetic retinopathy, kidney damage, heart disease, stroke, blindness, and nerve damage.2,5,6,8–10 Today, diabetes is diagnosed by measurement of blood glucose levels. It is widely agreed that frequent self-monitoring is an essential and central part of controlling diabetes, so diabetics need to supervise their glucose levels closely and measure them several times a day.3,7 The frequent monitoring of blood glucose is an essential part of diabetic management, as only the maintenance of a blood glucose level within the physiological range enables a diabetic to lead a healthy lifestyle by delaying or avoiding diabetic complications.1,2,10–12 Current blood glucose monitoring is accomplished by invasive methods to collect blood samples several times a day, such as a finger prick. However, frequent blood glucose monitoring was associated with patient pain, proper and complex analysis, and was time consuming. Frequent blood glucose monitoring is also expensive due to the number of required test strips, and it is liable to cause other diseases due to the penetration of the skin and discommodious operation.10,13,14 Therefore, a noninvasive technique may not only detect the blood glucose concentration duly, safely, and painlessly by patients themselves, but also be less expensive. For this purpose, a noninvasive method for blood glucose monitoring is highly desired.15 Most noninvasive systems are based on optical methods such as nearinfrared (NIR) spectroscopy, Raman spectroscopy, midinfrared (MIR) spectroscopy with an attenuated total reflection (ATR) prism, and polarimetry. Notably NIR and Raman spectroscopy have shown substantial promise in this regard.7,13,14 Near-infrared spectroscopy is supposed as a more promising technique among various methods for clinical blood glucose assay on the basis of its potential for rapid analysis and reagentless, nondestructive, and noninvasive measurements. As defined by the American Society of Testing and Materials, the NIR region of the electromagnetic spectrum spans wavelengths ranging from 780 to 2526 nm. The most prominent absorption bands of NIR are related to the overtones and combina-

0003-7028/14/6804-0428/0 Q 2014 Society for Applied Spectroscopy

APPLIED SPECTROSCOPY

tions of fundamental vibrations exhibited by –CH, –NH, –OH, and –SH functional groups.16,17 However, due to the relatively low absorbance of glucose, the NIR spectroscopy could be affected severely by strong absorption of other components in blood such as water, protein, and fats, and the variations of physiological status and measurement conditions at different times.7,13,18,19 Because Raman scattering of water is weak, and Raman spectroscopy usually provides sharper and less overlapped spectra compared, for instance, to NIR, Raman spectroscopy has the advantage of providing rich information about the molecular structure of the sample. Raman spectroscopy is based upon the inelastic scattering interactions between light and matter. These interactions produce spectral bands, formed by photons scattered at the frequencies of the molecular vibrations that provide valuable sensitive and specific information of the biochemical composition of the tissue.20 Recently, sophisticated data analysis techniques based on multivariate analysis have made it possible to exploit the full information content of Raman spectra and to draw conclusions about the chemical structure and composition of very complex systems such as detecting of glucose in serum, whole blood, and human aqueous humor.3,15,19,21–24 However, some limitations of Raman spectroscopy for biomedical applications have included issues such as weak scattering signals, long spectral acquisition times, fluorescence from biological samples, the instability of the laser intensity, and the photothermal damage to samples.22,25,26 Artificial plasma (AP) is extensively used as a plasma substitute for plasma volume expansion in managing the surgical, medical, and critically ill intensive care patients.27–29 In our research, the main effect of the physiological components, including water, protein, and fats, was simulated, respectively, by two kinds of artificial plasma, which were hydroxyethyl starch (HES) and gelatin solutions. Near-infrared and Raman spectroscopy were used to realize a noninvasive glucose assay in two kinds of artificial plasma solution samples. The two spectroscopic models were generated by performing partial least-squares (PLS) regression and validated for the determination of glucose, and the results show: (i) the two spectroscopic models established are robust, accurate, and repeatable and can be used for quantitative analysis of glucose, and (ii) the predictions of NIR spectroscopy were superior to that of Raman spectroscopy.

MATERIALS AND METHODS Samples and Reagents. Two kinds of artificial plasma were used in our research; the artificial plasma I (AP-1) was two different concentrations of hydroxyethyl starch (HES) solution (95% and 100% HES injection), and the artificial plasma II (AP-II) was two different concentrations of gelatin solution (95% and 100% gelatin injection). There are 100 samples for each kind of AP with glucose concentration range of 0.1  5.0 mg/mL, i.e., 10  500 mg/dL. Data Collection. The NIR spectra were recorded using a Bruker Matrix-F FT-NIR spectrometer (Bruker Optik, Ettlingen, Germany) equipped with a PbS detector,

sample cup, and a fiber optic probe. The system was operated by OPUS spectral acquisition and processing software (Bruker Optik, Ettlingen, Germany). The spectra were obtained at a resolution of 8 cm1 over a wavelength range of 12 000–4000 cm1 with 32 scans per spectrum, and air absorbance was recorded as the reference standard. The Raman spectra were obtained using a Metage OPAL Portable Raman System (ProRaman L-785, EVWAVE Optronics. Inc.) equipped with a thermoelectrically controlled CCD detector (cooled to 50 8C), stander laser excitation at 785 nm diode laser (power output up to 300–400 mW), sample cup, and a fiber optic probe. The system was operated by eFTIR spectral acquisition and processing software (Essential FTIR V3.00.047 from Operant LLC Licensed to MTG). Raman spectra were recorded at a spectral resolution of 4 cm1 over a wavelength range of 3200–90 cm1 with ten scans per spectrum. Data Processing. The intensity of the measurements at different wavenumbers (Raman shifts) can be correlated to the concentrations of the relevant components in the sample through a series of mathematical procedures. These processes involve multivariate statistical calculations such as multiple linear regression (MLR), principal component regression (PCR), the partial least squares (PLS), the artificial neural networks, and so on. Partial least squares was the most frequently used in these methods. Therefore, the NIR and Raman spectroscopic calibration models were constructed, respectively, by using PLS with the OPUS software.16,30

RESULTS AND DISCUSSION Spectral Features and Sample Set Selection. Figure 1 shows the original NIR and Raman spectra of 100 samples of AP-I and AP-II. Among the 100 samples for each kind of AP, 21 samples were selected randomly for the validation set, and the remaining 79 samples were for the calibration set. Due to the relatively low absorbance of glucose and effect of other components in the NIR and Raman spectra, the curves of spectra had very similar shapes, and all showed broad and overlapped peaks. Therefore, the NIR and Raman spectra of a higher concentration of glucose that was dissolved in AP-I and AP-II was collected and is shown in Fig. 2. Spectral Pretreatment Methods. To develop a robust model, different spectral pretreatment methods are often utilized to eliminate noise, baseline shift, and matrix background interference, and enhance the spectral features to extract the relevant information before PLS modeling. The OPUS software provides 11 kinds of important spectral pretreatment methods, including: untreated, constant offset elimination (COE), straight line subtraction (SLS), vector normalization (VN), min/ max normalization (MMN), multiplicative scatter correction (MSC), first derivative, second derivative, first derivative þ SLS, first derivative þ VN, and first derivative þ MSC. In the spectral pretreatment process, the PLS model is validated by cross-validation to assess the average predictive ability. The cross-validation method is a

APPLIED SPECTROSCOPY

429

FIG. 1. The NIR and Raman spectra of the 100 samples of AP-I and AP-II. (a) NIR spectra of AP-I, (b) NIR spectra of AP-II, (c) Raman spectra of AP-I, (d) Raman spectra of AP-II.

resampling technique based on the leave-one-out sample procedure. This means that the calibration is performed N times (N is the number of samples in the calibration dataset), each time leaving one sample out and testing the calibration equation on this single sample. The PLS methods must be computed for each number of components, and the optimum number of factors for calibration was selected based on the correlation coefficients of the calibration set (Rcal), the root mean square errors of cross-validation (RMSECV), and the residual predictive deviation (RPD). The model with the best prediction ability is usually selected by computing the root mean square error of validation (RMSEP) and the correlation coefficients of the validation set (Rval).16,31,32 The OPUS could give the value of the R, RMSECV, RPD, and RMSEP, and then the best models were chosen according to these evaluation parameters; the calibration models with highest R (both in calibration and validation) and RPD as well as lowest RMSECV and RMSEP were considered optimal. According to the above criteria, the best calibration models were generated based on min/max normalization (MMN) for the NIR and Raman models of AP-I, multiplicative scatter correction (MSC) for the NIR model of AP-II, and constant offset elimination (COE) for the Raman model of AP-II, as shown in Table I. Spectral Region and Factors. In the complex system, it was difficult to use a classical univariate calibration method for quantitative analysis. In addition, one or

430

Volume 68, Number 4, 2014

several wavelengths related to glucose content could not be found because of the interferences of the other components on the NIR and Raman spectra, so the spectral region rich in chemical information was used usually to establish the calibration model. The partial least squares (PLS) method of the OPUS is a powerful multivariate calibration method in recommending the appropriate spectral region. Table I shows the appropriate spectral region for the NIR and Raman models. In the same spectral region and spectral pretreatment method, the number of PLS factors (F) would directly affect R, RPD, RMSECV, and RMSEP. Not enough information would be obtained from the spectrum when F is too small, and the actual information is misrepresented or ‘‘over fitted’’ when F is too big. The OPUS was successful in recommending the ideal F according to the criteria of the best calibration. As shown in Table I, the ideal F for the NIR model of AP-I, the NIR model of AP-II, the Raman model of AP-I, and the Raman model of AP-I was 18, 16, 6, and 6, respectively. With partial least squares (PLS) regression, the calibration models were optimized individually by considering spectral range, spectral pretreatment methods, and number of model factors. The values of the R, RMSECV, RPD, and RMSEP for the best calibration models are shown in Table I. Evaluation and Validation for the Optimal NearInfrared (NIR) and Raman Models. Some figures of merit were taken into account in the present study in

FIG. 2. The NIR and Raman spectra of water (1, green curve), AP-I or AP-II (2, blue curve), and 100 mg/ml glucose dissolved in AP-I or AP-II (3, red curve). (a) NIR spectra of AP-I, (b) NIR spectra of AP-II, (c) Raman spectra of AP-I, (d) Raman spectra of AP-II.

order to evaluate and validate the prediction ability of two spectroscopic models as follows. Accuracy. This parameter reports the closeness of agreement between the reference value and the value predicted by the calibration model. In chemometrics, this is generally expressed as RMSEP and R of validation (Rval). In addition, the statistical test and recovery tested were carried out to evaluate and validate the optimal models. The above optimized models were used to predict, respectively, the content of 21 samples in the validation sets of AP-I and AP-II. As shown in Table I, the RMSEP for the NIR model of AP-I, the NIR model of AP-II, the Raman model of AP-I, and the Raman model of AP-II was 0.061, 0.046, 0.173, and 0.156, respectively, and the R of validation (Rval) was 99.78%, 99.88%, 98.13%, and

98.59%, respectively. Compared to Raman spectroscopy, the performance of NIR spectroscopy was much better, with lower RMSEP and higher Rval. The T and F tests for the predicted result of the validation set indicated that the accuracy was satisfactory with a significant level of 0.05. The accuracy was also tested as recovery in three known concentration levels (approximately 1.0, 2.0, and 3.0 mg/ml). Then nine samples of the three levels for external validation were analyzed, respectively, by the optimal models. The average recoveries were calculated by the following equation: Recoveryð%Þ ¼

concentration analyzed 3 100% reference concentration

ð1Þ

TABLE I. Optimized parameters used by the PLS model and calibration and validation results for estimation by NIR and Raman spectroscopy. Spectral pretreatment method

Factor

Spectrum region (cm1)

Rcal (%)

RMSECV

RPD

RMSEP

Rval (%)

NIR model of AP-I NIR model of AP-II Raman model of AP-I

Min/max normalization Multiplicative scatter correction Min/max normalization

18 16 6

99.15 99.55 92.22

0.128 0.094 0.491

11 15 2.6

0.061 0.046 0.173

99.78 99.88 98.13

Raman model of AP-II

Constant offset elimination

11995.6–5446.2 11995.6–5446.2 3220.6–1969 1657.8–406.3 3220.6–2594.1 2282–1969.5 1344.7–406.1

93.29

0.470

2.7

0.156

98.59

6

APPLIED SPECTROSCOPY

431

TABLE II. The recovery and precision test of glucose assay by NIR and Raman spectroscopy. Recovery test

NIR model of AP-I Low-concentration Mid-concentration High-concentration NIR model of AP-II Low-concentration Mid-concentration High-concentration Raman model of AP-I Low-concentration Mid-concentration High-concentration Raman model of AP-II Low-concentration Mid-concentration High-concentration

Precision test

Recovery (%)

RSD (%)

Repeatability (%)

Intraday variability (%)

Interday variability (%)

104.80 100.27 96.64

1.22 2.26 0.32

0.84 0.64 0.51

1.22 2.68 1.03

2.61 2.28 1.89

98.50 100.66 100.29

2.93 3.78 1.14

0.84 0.98 0.52

1.87 1.22 2.08

2.52 2.23 1.65

103.56 101.87 99.26

2.79 3.78 2.49

2.19 2.09 1.12

3.56 0.77 2.82

3.81 2.47 2.54

98.78 102.81 100.35

3.78 2.19 2.33

3.12 1.56 1.60

3.54 2.34 1.88

3.49 2.64 2.31

Relative  standard deviationð%Þ  standard deviation 3 100% ¼ mean

CONCLUSION ð2Þ

The recovery test results for three levels are listed in Table II. In the two spectroscopic models, the overall recovery was between 95% and 105% with the relative standard deviation (RSD) less than 4.0% for the analytes. These studies demonstrated that the two models had good accuracy, and the NIR model showed slightly better performance than the Raman model. Precision. It represents the degree of scatter between a series of measurements for the same sample. It is usually expressed as a relative standard deviation (RSD) of a series of measurements. The precision levels evaluated in this work were (i) repeatability: it was determined by three replicate measurements for every single sample of three known concentration levels (approximately 1.0, 2.0, and 3.0 mg/ml), respectively; (ii) intra- and interday variability: The intraday variability was determined, respectively, by nine samples of the three levels through the optimal models within a day, while for the interday variability test, the solution was examined in duplicate for another day. As shown in Table II, the repeatability of the developed models had good performance with the RSD less than 1.0% for the NIR models and less than 3.5% for the Raman models, so the repeatability of the NIR models showed much better performance. The intra- and interday variability was less than 3.0% for the NIR models, while it was less than 4.0% for the Raman models, so the NIR model showed slightly better performance in the intra- and interday variability test. As shown in Tables I and II, the evaluation and validation results show that the two spectroscopic models established are robust, accurate, and repeatable and can be used for quantitative analysis of glucose. Compared to Raman spectroscopy, the performance of NIR spectroscopy was much better in the accuracy and precision test.

432

Volume 68, Number 4, 2014

In this study, the NIR and Raman spectroscopy provided robust, accurate, repeatable, and rapid analysis for glucose in two kinds of artificial plasma. The two spectroscopic models established can be used for noninvasive measurement of glucose in artificial plasma. As the NIR spectroscopy shows better performance and extensive potential in noninvasive measurement of glucose, it will be carried out in whole blood and skin tissue in vitro or in vivo in our future study. 1. P. Zimmet, K. Alberti, J. Shaw. ‘‘Global and Societal Implications of the Diabetes Epidemic’’. Nature. 2001. 414(6865): 782-787. 2. S.K. Vashist. ‘‘Non-Invasive Glucose Monitoring Technology in Diabetes Management: A Review’’. Anal. Chim. Acta. 2012. 750: 1627. 3. J. Shao, M. Lin, Y. Li, X. Li, J. Liu, J. Liang, H. Yao. ‘‘In Vivo Blood Glucose Quantification Using Raman Spectroscopy’’. PLoS One. 2012. 7(10): e48127. 4. M.R. Stearne, S.L. Palmer, M.S. Hammersley, S.L. Franklin, R.S. Spivey, J.C. Levy, C.R. Tidy, N.J. Bell, J. Steemson, B.A. Barrow, R. Coster, K. Waring, J. Nolan, E. Truscott, N. Walravens, L. Cook, H. Lampard, C. Merle, P. Parker, J. McVittie, I. Draisey, L.E. Murchison, A.H.E. Brunt, M.J. Williams, D.W. Pearson, X.M.P. Petrie, M.E.J. Lean, D. Walmsley, M.J. Lyall, E. Christie, J. Church, E. Thomson, A. Farrow, J.M. Stowers, M. Stowers, K. McHardy, N. Patterson, A.D. Wright, N.A. Levi, A.C.I. Shearer, R.J.W. Thompson, G. Taylor, S. Rayton, M. Bradbury, A. Glover, A. Smyth-Osbourne, C. Parkes, J. Graham, P. England, S. Gyde, C. Eagle, B. Chakrabarti, J. Smith, J. Sherwell, E.M. Kohner, A. Dornhurst, M.C. Doddridge, M. Dumskyj, S. Walji, P. Sharp, M. Sleightholm, G. Vanterpool, C. Rose, G. Frost, M. Roseblade, S. Elliott, S. Forrester, M. Foster, K. Myers, R. Chapman, J.R. Hayes, R.W. Henry, M.S. Featherston, G.P.R. Archbold, M. Copeland, R. Harper, I. Richardson, S. Martin, M. Foster, H.A. Davison, D.R. Hadden, L. Kennedy, A.B. Atkinson, A.M. Culbert, C. Hegan, H. Tennet, N. Webb, I. Robinson, J. Holmes, M. Foster, P.M. Bell, D.R. McCance, J. Rutherford, S. Nesbitt, A.S. Spathis, S. Hyer, M.E. Nanson, L.M. James, J.M. Tyrell, C. Davis, P. Strugnell, M. Booth, H. Petrie, D. Clark, B. Rice, S. Hulland, J.L. Barron, J.S. Yudkin, B.J. Gould, J. Singer, A. Badenock, S. Walji, M. Eckert, K. Alibhai, E. Marriot, C. Cox, R. Price, M. Fernandez, A. Ryle, S. Clarke, G. Wallace, E. Mehmed, S. MacFarlane, R.H. Greenwood, J. Wilson, M.J. Denholm, R.C. Temple, K. Whitfield, F. Johnson, C. Munroe, S. Gorick, E. Duckworth, M. Flatman, S. Rainbow, L.J. Borthwick, D.J. Wheatcroft, R.J. Seaman, R.A. Christie, W. Wheatcroft, P. Musk, J. White, S. McDougal, M. Bond, P. Raniga, R.W. Newton, R.T. Jung, C.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

Roxburgh, B. Kilgallon, L. Dick, M. Foster, N. Waugh, S. Kilby, A. Ellingford, J. Burns, C.V. Fox, M.C. Holloway, H.M. Coghill, N. Hein, A. Fox, W. Cowan, M. Richard, K. Quested, S.J. Evans, R.B. Paisey, N.P.R. Brown, A.J. Tucker, R. Paisey, F. Garrett, J. Hogg, P. Park, K. Williams, P. Harvey, R. Wilcocks, S. Mason, J. Frost, C. Warren, P. Rocket, L. Bower, J.M. Roland, D.J. Brown, J. Youens, K. StantonKing, H. Mungall, V. Ball, W. Maddison, D. Donnelly, S. King, P. Griffin, S. Smith, S. Church, G. Dunn, A. Wilson, K. Palmer, P.M. Brown, D. Humphriss, A.J.M. Davidson, R. Rose, L. Armistead, S. Townsend, P. Poon, I.D.A. Peacock, N.J.C. Culverwell, M.H. Charlton, B.P.S. Connolly, J. Peacock, J. Barrett, J. Wain, W. Beeston, G. King, P.G. Hill, A.J.M. Boulton, A.M. Robertson, V. Katoulis, A. Olukoga, H. McDonald, S. Kumar, F. Abouaesha, B. Abuaisha, E.A. Knowles, S. Higgins, J. Booker, J. Sunter, K. Breislin, R. Parker, P. Raval, J. Curwell, H. Davenport, G. Shawcross, A. Prest, J. Grey, H. Cole, C. Sereviratne, R.J. Young, T.L. Dornan, J.R. Clyne, M. Gibson, I. O’Connell, L.M. Wong, S.J. Wilson, K.L. Wright, C. Wallace, D. McDowell, A.C. Burden, E.M. Sellen, R. Gregory, M. Roshan, N. Vaghela, M. Burden, C. Sherriff, J. Clarke, J. Grenfell, J.E. Tooke, K. MacLeod, C. Searnark, M. Rammell, C. Pym, J. Stockman, C. Yeo, J. Piper, L. Leighton, E. Green, M. Hoyle, K. Jones, A. Hudson, A. James, A. Shore, A. Higham, B. Martin. ‘‘Tight Blood Pressure Control and Risk of Macrovascular and Microvascular Complications in Type 2 Diabetes: UKPDS 38’’. BMJ. 1998. 317(7160): 703-713. C.M. Steppan, S.T. Bailey, S. Bhat, E.J. Brown, R.R. Banerjee, C.M. Wright, H.R. Patel, R.S. Ahima, M.A. Lazar. ‘‘The Hormone Resistin Links Obesity to Diabetes’’. Nature. 2001. 409(6818): 307-312. M.W. Aslam, Z. Zhu, A.K. Nandi. ‘‘Feature Generation Using Genetic Programming with Comparative Partner Selection for Diabetes Classification’’. Expert. Syst. Appl. 2013. 40(13): 5402-5412. R.S. Parker, F.J. Doyle, N.A. Peppas. ‘‘The Intravenous Route to Blood Glucose Control’’. IEEE Eng. Med. Biol. Mag. 2001. 20(1): 6573. S.E. Capes, D. Hunt, K. Malmberg, H.C. Gerstein. ‘‘Stress Hyperglycaemia and Increased Risk of Death after Myocardial Infarction in Patients with and without Diabetes: A Systematic Overview’’. Lancet. 2000. 355(9206): 773-778. I.M. Stratton, A.I. Adler, H.A.W. Neil, D.R. Matthews, S.E. Manley, C.A. Cull, D. Hadden, R.C. Turner, R.R. Holman. ‘‘Association of Glycaemia with Macrovascular and Microvascular Complications of Type 2 Diabetes (UKPDS 35): Prospective Observational Study’’. BMJ. 2000. 321(7258): 405-412. J.D. Newman, A.P.F. Turner. ‘‘Home Blood Glucose Biosensors: A Commercial Perspective’’. Biosens. Bioelectron. 2005. 20(12): 24352453. ADVANCE Collaborative Group, A. Patel, S. MacMahon, J. Chalmers, B. Neal, L. Billot, M. Woodward, M. Marre, M. Cooper, P. Glasziou, D. Grobbee, P. Hamet, S. Harrap, S. Heller, L. Liu, G. Mancia, C.E. Mogensen, C. Pan, N. Poulter, A. Rodgers, B. Williams, S. Bompoint, B.E. de Galan, R. Joshi, F. Travert. ‘‘Intensive Blood Glucose Control and Vascular Outcomes in Patients with Type 2 Diabetes’’. N. Engl. J. Med. 2008. 358(24): 2560-2572. U.K. Prospective Diabetes Study Group. ‘‘Effect of Intensive BloodGlucose Control with Metformin on Complications in Overweight Patients with Type 2 Diabetes (UKPDS 34)’’. Lancet. 1998. 352(9131): 854-865. Y. Yang, Z.Z. Shi, C.X. Li, W.-L. Chen, K.-X. Xu. ‘‘Simulation and Validation of the Radial Reference Point in Non-Invasive Blood Glucose Sensing by NIR’’. Nanotechnol. Precis. Eng. 2010. 8(02): 6. Z.-M. Chuah, R. Paramesran, K. Thambiratnam, S.-C. Poh. ‘‘A TwoLevel Partial Least Squares System for Non-Invasive Blood Glucose Concentration Prediction’’. Chemom. Intell. Lab. Syst. 2010. 104(2): 347-351. K. Maruo, M. Tsurugi, J. Chin, T. Ota, H. Arimoto, Y. Yamada, M. Tamura, M. Ishii, Y. Ozaki. ‘‘Noninvasive Blood Glucose Assay

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29. 30.

31.

32.

Using a Newly Developed Near-Infrared System’’. IEEE J. Sel. Top. Quant. 2003. 9(2): 322-330. J. Xue, C. Wu, L. Wang, S. Jiang, G. Huang, J. Zhang, S. Wen, L. Ye. ‘‘Dynamic Prediction Models for Alkaloid Content Using NIR Technology for the Study and Online Analysis of Parching in Areca Seed’’. Food Chem. 2011. 126(2): 725-730. J.T. Xue, C.J. Wu, L.L. Wang, J.L. Zhang, G. Huang, S. Jiang, S.L. Wen, Y. Liang, R.B. Chao, C. Chen. ‘‘Identification of the Types of Parched Areca Seed by Near Infrared Spectroscopy’’. Asian J. Chem. 2012. 24(2): 822-826. Q.-B. Li, L.-N. Li, G.-J. Zhang. ‘‘A Nonlinear Model for Calibration of Blood Glucose Noninvasive Measurement Using Near Infrared Spectroscopy’’. Infrared Phys. Technol. 2010. 53(5): 410-417. A. Tura, A. Maran, G. Pacini. ‘‘Non-Invasive Glucose Monitoring: Assessment of Technologies and Devices According to Quantitative Criteria’’. Diabetes Res. Clin. Pract. 2007. 77(1): 16-40. E.E. Rossi, A.L. Pinheiro, O.C. Baltatu, M.T. Pacheco, L. Silveira Jr. ‘‘Differential Diagnosis Between Experimental Endophthalmitis and Uveitis in Vitreous with Raman Spectroscopy and Principal Components Analysis’’. J. Photochem. Photobiol. B. 2012. 107: 7378. J.M. Yuen, N.C. Shah, J.T. Walsh Jr., , M.R. Glucksberg, R.P. Van Duyne. ‘‘Transcutaneous Glucose Sensing by Surface-Enhanced Spatially Offset Raman Spectroscopy in a Rat Model’’. Anal. Chem. 2010. 82(20): 8382-8385. I. Barman, C.-R. Kong, N.C. Dingari, R.R. Dasari, M.S. Feld. ‘‘Development of Robust Calibration Models Using Support Vector Machines for Spectroscopic Monitoring of Blood Glucose’’. Anal. Chem. 2010. 82(23): 9719-9726. K. Kneipp, H. Kneipp, I. Itzkan, R.R. Dasari, M.S. Feld. ‘‘Ultrasensitive Chemical Analysis by Raman Spectroscopy’’. Chem. Rev. 1999. 99(10): 2957-2976. J.L. Lambert, C.C. Pelletier, M. Borchert. ‘‘Glucose Determination in Human Aqueous Humor with Raman Spectroscopy’’. 2005. 10(3): 031110. K. Shin, H. Chung. ‘‘Wide Area Coverage Raman Spectroscopy for Reliable Quantitative Analysis and Its Applications’’. Analyst. 2013. 138(12): 3335-3346. D.I. Ellis, D.P. Cowcher, L. Ashton, S. O’Hagan, R. Goodacre. ‘‘Illuminating Disease and Enlightening Biomedicine: Raman Spectroscopy as a Diagnostic Tool’’. Analyst. 2013. 138(14): 38713884. C. Fenger-Eriksen, E. Anker-Moller, J. Heslop, J. Ingerslev, B. Sorensen. ‘‘Thrombelastographic Whole Blood Clot Formation after Ex Vivo Addition of Plasma Substitutes: Improvements of the Induced Coagulopathy with Fibrinogen Concentrate’’. Brit. J. Anaest. 2005. 94(3): 324-329. S.N. Mardel, F.M. Saunders, H. Allen, G. Menezes, C.M. Edwards, L. Ollerenshaw, D. Baddeley, A. Kennedy, R.M. Ibbotson. ‘‘Reduced Quality Of Clot Formation with Gelatin-Based Plasma Substitutes’’. Brit. J. Anaest. 1998. 80(2): 204-207. J. Boldt, S. Suttner. ‘‘Plasma Substitutes’’. Minerva Anestesiol. 2005. 71(12): 741-758. N. Heigl, A. Greiderer, C.H. Petter, O. Kolomiets, H.W. Siesler, M. Ulbricht, G.K. Bonn, C.W. Huck. ‘‘Simultaneous Determination of the Micro-, Meso-, and Macropore Size Fractions of Porous Polymers by a Combined Use of Fourier Transform Near-Infrared Diffuse Reflection Spectroscopy and Multivariate Techniques’’. Anal. Chem. 2008. 80(22): 8493-8500. X. Niu, Z. Zhao, K. Jia, X. Li. ‘‘A Feasibility Study on Quantitative Analysis of Glucose and Fructose in Lotus Root Powder by FT-NIR Spectroscopy and Chemometrics’’. Food. Chem. 2012. 133(2): 592597. S. Sivakesava, J. Irudayaraj, D. Ali. ‘‘Simultaneous Determination of Multiple Components in Lactic Acid Fermentation Using FT-MIR, NIR, and FT-Raman Spectroscopic Techniques’’. Process. Biochem. 2001. 37(4): 371-378.

APPLIED SPECTROSCOPY

433