|
1.IntroductionNear-infrared (NIR) spectroscopy is one of the most promising methods for noninvasive blood glucose concentration (BGC) measurement.1–3 However, its application in long-term in vivo BGC monitoring remains a challenge due to unsatisfactory accuracy.4–6 The established spectral prediction model based on BGC values measured using the oral glucose tolerance test method is commonly considered “time-limited” because it would not be available after a day.4–6 Generally, background disturbances during long-term monitoring, which may involve instrument drift, change of surrounding temperature, variations of man–machine interfaces, physiological fluctuation in the body, etc., would markedly change the spectra and lead to an unacceptable deviation for predicted BGC values.3,7 Except for unavoidable disturbances caused by physiological fluctuation, the other disturbances related to instrument or man–machine interfaces should be well controlled and reduced. Some advanced devices with reference optical paths were applied with the goal of eliminating the drift of signals from instruments.8,9 Some probes designed with attached sensors have also been applied because they can precisely monitor the location on the skin and reproduce the skin contact for repeated tests.10 However, there is still random fluctuation of incident light intensity in these cases because the incident angle or the surface reflectance of light varies randomly. We are planning to develop an effective mathematical correction method to try to greatly reduce the drift related to varying incident light intensities, which may be caused by the randomly varying skin-probe contact or drift of the light source. Our correction is based on using a reference signal from the subject itself, and the reference is considered a good index for monitoring the variations related to skin-probe contact as well as the physiological skin conditions.11 This reference-based correction and the subsequent measurements might yield higher accuracy in long-term BGC monitoring.12 This correction method can be called an in vivo reference-based correction approach (VRCA). Moreover, the reference can be set in various ways, such as a time-based,13 position-based,13 or wavelength-based reference.14 In recent years, a position-based VRCA has been considered promising in BGC measurements because it can provide satisfactory glucose absorption information with a sufficiently long equivalent optical path length.15,16 In addition, the introduction of a floating reference position (FRP), which involves a special source–detector (SD) separation, has been identified as an excellent position reference. At an FRP, the change in diffuse reflection of the light intensity signal caused by disturbances can be directly monitored because it is sensitive to disturbances but insensitive to BGC variations.16–19 However, the correction using FRP is also challenging because the drifts from different positions, i.e., SD separations, are completely different from their reference, the FRP. The relationship between the drifts from the focused SD separations and the FRP should first be determined. Zhang19 estimated the ratios between the drifts from the FRP and two other SD separations when the drifts occurred as the light source was set on different output powers, and then the correction was performed using the ratios. However, the empirical ratios are only available for the two SD separations and wavelengths used. It is difficult to reference other SD separations and wavelengths. In this paper, we report that the relative changes in diffuse reflection intensity are approximately linear with SD separations when either the incident light intensity or glucose concentration varies. In previous investigations, the absolute changes in diffuse reflection intensity from different SD separations were considered totally different and exponential-like. After using the relative approach, we recognized that the linearity of the drift could be a very helpful indicator of the relationship between the variations from different positions including the FRP. We could even estimate all possible values of the relative change of diffuse light intensity for all positions by performing spectral data recording at only a few positions. Furthermore, we found that the relative changes seem to be similar for the different positions where only incident light intensity varies. Therefore, the ratio between the drifts from different SD separations and FRP can be easily acquired using an equation we developed. The correction can easily be performed for all wavelengths from 1100 to 1400 nm as well as for SD separations. 2.Materials and Methods2.1.Experimental SystemA supercontinuum source (SC46, YSL photonics, China) with a broadband from 600 to 1700 nm is used as light source in the custom-built NIR spectrometer. As shown in Fig. 1, the system is equipped with a noncollinear acousto-optic tunable filter (TEAF10-1.0-1.8-S, Brimrose Company, Maryland) as the light splitting element to provide a spectral response range from 1100 to 1400 nm (9091 to ) with a resolution of 2 nm, an InGaAs infrared detector (G12181-210K, Hamamatsu Photonics K.K., Japan), and a 16-bit data acquisition board (PCI-6220, National Instrument Inc.). The fiber used in the experiment is multimode fiber, its core diameter is , and its numerical aperture is 0.22. 2.2.SamplesNIR diffuse reflection spectra were recorded for the scattering media of 10% Intralipid solution20,21 with different glucose concentrations, which are from 0 to with an interval of . 2.3.MeasurementsTwenty-two detection distances of SD separations were set from 0.47 to 3.095 mm with an interval of 0.125 mm. The surrounding temperature was controlled at 25°C. All samples were measured 5 times within 48 h, and the samples were tested in a random order every time. The samples were placed on a magnetic plate and stirred for 1 min and then allowed to settle for 3 min before recording. The test time was arranged to start at 9:16 am and 9:16 pm of the first day, 9:15 am and 9:15 pm of the second day, and 9:16 am of the third day. We call the five times tests first, second, third, fourth, fifth tests, respectively. Measuring all SD separations and wavelengths for one sample took and all 11 samples took 10 min. Every separate test from the five tests can be considered as a short-term test, where the can demonstrate caused by glucose. But between the five tests, which lasted for 48 h, the for a same sample can demonstrate caused by light source drift. 2.4.Defining Relative Changes in Diffuse Reflection Light IntensityFigure 2 shows the diffused light in a scattering medium, where is the incident light intensity, and is the diffuse reflection light intensity. The relative variation of diffuse reflection light is defined as in Eq. (1), where is the diffuse reflection light intensity and is the variation of as referenced to an initial state of the scattering medium. denotes the diffuse light intensity for reference. can be induced by glucose or disturbances and denote the relative changes caused by glucose concentration and the incident light intensity’s drift, respectively. is for both causes. When the variations are not too big, the can be decomposed to the two parts, i.e., . and denote the variations of caused by glucose concentration and the incident light intensity’s drift, respectively. Therefore, the equation of would work, too.2.5.Linearity of Relative Changes in Diffuse Reflection Light IntensityUsing diffuse equation (DE), i.e., the first-order approximate radiative transfer equation (P1-RTE),22 and Monte-Carlo (MC) simulation,23 the linearity of can be demonstrated using their results. The diffused light intensity can be calculated using Eq. (2), which is solved for a semi-infinite medium by setting an extrapolative boundary where is the SD separation, is the effective attenuation coefficient, defined as . is the absorption coefficient of the medium; is the reduced scattering coefficient defined as , where is the anisotropy factor, and is the scattering coefficient; is the diffusion coefficient as . , , , and .In the glucose measurement, can be calculated after the diffuse light intensity values with different glucose concentrations are estimated by using Eq. (2) or an MC stimulation program. The 10% Intralipid solution’s parameters24 , , , and used in the equation and MC simulation are shown in Table 1. The variations of the parameters caused by glucose can be determined according to the equations in Kohl’s papers.25,26 The variation of the absorption coefficient induced by the glucose concentration can be expressed as25 , where and are the molar extinction coefficients of glucose and water, respectively, and is the displacement factor between glucose and water. The variations of the refractive index and anisotropic factor induced by the glucose concentration are expressed as25 , . For Intralipid solutions, the variation of the scattering coefficient induced by the glucose concentration is26 , where , and is the wavelength (nm). Table 1The parameters of 10% Intralipid used in P1-RTE and MC at 1100 to 1400 nm.
Figures 3 and 4 show the results of for the 1100-nm diffuse light from the SD separations of 0 to 3.5 mm. The in Figs. 3 and 4 can show an approximate linearity to SD separation. Glucose would induce a simultaneous change in and of the tested medium, and the two factors would induce an approximately linear change in to SD separation. The linearity caused by glucose could be a comprehensive result. It also can be found that the diffuse light intensity around the FRP would be relative stable and insensitive to the change of glucose concentration. The values of FRP appear different in Figs. 3 and 4 because of the differences of the two theoretical practices using DE and MC simulations. 3.Experimental Results and Calculations3.1.Result of in One Short-Term TestThe FRP is found out by the phantom experiment. The relative variations were calculated for the samples with the glucose concentration of 200 to using Eq. (1) when the is from the sample without glucose. Figure 5 shows some typical results of . Figure 6 shows the FRPs of 1100 to 1400 nm by MC when changes . The location of FRP is relatively stable when the scattering properties of the medium change a little. The FRPs of MC are different with the experiments because MC is based on the assumption of a semi-infinite medium, while it is actually closer to an infinite medium in the experiments. 3.2.Result of with Instrument Drift During 48 hUsing the diffuse reflection light intensity results from the first test as the value, the results can be acquired for all samples to demonstrate their drifts in the second to fifth tests. For every sample, the primarily shows the influence from the instrument drift. Some typical results are shown in Fig. 7 and the from different SD separations appear approximately similar. It can, therefore, be assumed that there is an equivalent for an arbitrary pair of SD separations of 0.47 to 3.095 mm as shown in where and are two different SD separations.Based on Eq. (3), the light source intensity drift can be reduced by differential calculation between the of two SD separations. As shown in The glucose-related information in after differential calculation is a combination of the glucose information from and . If one of the SD separations is the FRP, does not contain any glucose information, i.e., . It can be derived that would be the corrected at , i.e., . So, the correction can be described as is any SD separation where the spectral data need to be corrected. Equation (5) can also be described as Therefore, using Eq. (6), the change of light intensity at any position () can be estimated by the drift at the FRP () when it takes the FRP as the reference position. It should be noted that the characteristics of the synchronic drift might not occur with weaker diffuse light signals, which are reflected at the distances far from the light source. For the wavelengths where the medium has stronger absorption, the available SD separations would be very small. 3.3.Correcting based on the FRPFrom Eq. (3), the FRP can directly indicate the drift even for all available positions. When the measurement is only disturbed by the drift of incident light intensity or glucose concentration variation, the drifted can be corrected for all SD separations when the deducts the drift at , as shown in Eq. (5). Figure 8(a) shows the results, including and , for glucose solution samples referencing to the solution without glucose. The corrected result is shown in Fig. 8(b). Using the correction method based on FRP, the results from all samples and all tests are corrected. Figure 9 shows the correction results for some typical samples in the last test, which started 48 h after the first test. And all the samples’ are calculated using Eq. (1), in that the is the reference light intensity measured in the solution without glucose. According to Eq. (4), the light source drift can be reduced by differential calculation, . But the result related to the two positions , will be different from any result of a single position. Figure 10 shows two results after correction. As can be seen, they are different from the short-term test shown in Fig. 5. In Fig. 10, the horizontal axis shows that , and are the reference positions. In Fig. 10(a), is the near light source, , and FRP is at 1.9 mm. In Fig. 10(b), is far from the light source, . As can be seen, choosing far from the light source does not contribute to a better result due to the increase of noise after differential calculation, although the glucose information increases at the same time. The selection of and could deeply influence the result of absolute change of light intensity . The detection limit can be evaluated by the comparison of and the level of noise. 3.4.Correction Results of all Spectra at 1100 to 1400 nmThe diffuse reflection spectra in the wavelength ranging from 1100 to 1400 nm are corrected using the reference position of FRP. Similarly, the absolute light intensity can be corrected using the corrected and its initial value. Figure 11 shows the absolute light intensity results for from 1100 to 1400 nm before and after correction for the SD separation of 0.595 mm. The coefficients of variation (CVs) of the spectra from the original and corrected results are compared in Fig. 12. It is clear that the CVs improved from 0.01 to 0.025 before correction to less than 0.004 after correction. 3.5.Modeling and Predicting Glucose Concentrations Using Corrected SpectraPartial least square and leave-one-out cross validation are used for modeling and predicting the glucose concentrations in 10% Intralipid. It is necessary to use the spectra from those positions, which are far away from the reference FRP, because their signals should be sensitive to glucose. For the wavelengths of 1100 to 1400 nm, the FRPs distribute between 1.2 and 2.0 mm. Therefore, the SD separations of 1.2 to 2.0 mm are not available. The data from greater than 3.2 mm could not be used due to the weak signals at some wavelengths. By randomly choosing spectra from the five tests for every sample, we can obtain a set of data to test the prediction with the disturbance of instrument drift. Using these data, the prediction has an unsatisfactory root mean square error of cross validation (RMSECV) result in the cross validation, as shown in the first RMSECV column result of Table 2. When all of the spectra are corrected using our approach, the RMSECV values are closer to the best values obtained from the spectra recorded in one test, which takes and can be called “short-term monitoring.” The average RMSECV values of the five short-term tests are listed in the third RMSECV column of Table 2. Moreover, the SD separations, which are near the FRP, should not be used for the glucose measurement, since there the signals got a worse prediction result due to being insensitive to the glucose concentration’s variation. Table 2RMSECV results of the prediction using the reference position of FRP.
Table 3 shows the RMSECVs of prediction after differential measurement between two SD separations. Choose as the detection position, and the referenced positions are the SD separations that are less than FRP, near FRP, and greater than FRP, respectively. For the wavelengths of 1100 to 1400 nm, FRP is between 1.2 and 2.0 mm (based on short-term experiment). The result shows that if is less than FRP, the differential result could decrease the glucose-related information because the two positions are too close to each other, and the prediction error is bigger than the result only using one position of . If is near to FRP, glucose information would be lost less and the RMSECVs using two positions are similar to the result using only one position of . For the far away from the FRP, the noise from greatly increases and the RMSECVs decrease after differential of the two positions. Table 3RMSECV results of the prediction using other positions as reference.
In Tables 2 and 3, the prediction of glucose after using the reference position of the accurate FRPs or of the positions that are near to FRP show good results, which is closer to the result from the short-term test. This is because the spectral data collected near the FRP carry little glucose-related information, so the corrected may reserve most original glucose information at . The prediction accuracy of selecting accurate FRPs or the positions near FRP as reference are similar. In real use, we suggest the FRPs of the subject should be measured first and then select the position that is easy to measure near the FRP. But the correction results using an accurate FRP have advantages of building a model because the corrected spectra and original spectra can be used for mutual prediction. 4.ConclusionsIn this paper, we present an FRP-based correction method for NIR diffuse reflection spectra, which we verified as an effective approach to reduce the drift caused by the variation of incident light intensity. This correction will be helpful for improving the stability of spectra during long-term BGC monitoring. The correction can be easily performed for all spectra at required positions when their relative changes of diffuse light intensity can be estimated by an FRP because changes in these relative variations are approximately linear with SD separations. Moreover, the correction can also be used to reduce the drifts from other sources of disturbance because their relative drifts in incident light intensity may yield similar characteristics of linearity. As shown in Figs. 2, 3, and 5, it is clear that glucose could also induce a linear change with a range of positions. Indeed, this characteristic of linearity is possessed by many components in solution or in blood when the available positions are limited to a reasonable range. We are also planning to reduce some mixed disturbances that linearly affect the relative change of diffuse light intensity by decomposing these mixed lines. In this paper, we made a preliminary successful attempt even though only the line of the drift from incident light is deduced for every signal. The correction method presented in this paper is significant for 48-h glucose concentration monitoring. The analysis method in this paper can also be referenced to eliminate other disturbances in a linear fashion and obtain high-quality spectral signals. AcknowledgmentsThis work was supported by the National Natural Science Foundation of China (No. 81471698), the National High Technology Research and Development Program of China (“863” Program, Nos. 2012AA022602 and 2012YQ090194), and the Project of “111” (No. B07014). ReferencesH. M. Heise et al.,
“Noninvasive blood glucose sensors based on near-infrared spectroscopy,”
Artif. Organs, 18
(6), 439
–447
(1994). http://dx.doi.org/10.1111/aor.1994.18.issue-6 ARORD7 0160-564X Google Scholar
S. F. Malin et al.,
“Noninvasive prediction of glucose by near-infrared diffuse reflectance spectroscopy,”
Clin. Chem., 45
(9), 1651
–1658
(1999). Google Scholar
K. Maruo et al.,
“In vivo noninvasive measurement of blood glucose by near-infrared diffuse-reflectance spectroscopy,”
Appl. Spectrosc., 57
(10), 1236
–1244
(2003). http://dx.doi.org/10.1366/000370203769699090 APSPA4 0003-7028 Google Scholar
J. Yadav et al.,
“Prospects and limitations of non-invasive blood glucose monitoring using near-infrared spectroscopy,”
Biomed. Signal Process. Control, 18 214
–227
(2015). http://dx.doi.org/10.1016/j.bspc.2015.01.005 Google Scholar
M. Goodarzi et al.,
“Multivariate calibration of NIR spectroscopic sensors for continuous glucose monitoring,”
TrAC, Trends Anal. Chem., 67 147
–158
(2015). http://dx.doi.org/10.1016/j.trac.2014.12.005 Google Scholar
I. Barman et al.,
“Development of robust calibration models using support vector machines for spectroscopic monitoring of blood glucose,”
Anal. Chem., 82
(23), 9719
–9726
(2010). http://dx.doi.org/10.1021/ac101754n ANCHAM 0003-2700 Google Scholar
K. Maruo et al.,
“New methodology to obtain a calibration model for noninvasive near-infrared blood glucose monitoring,”
Appl. Spectrosc., 60
(4), 441
–449
(2006). http://dx.doi.org/10.1366/000370206776593780 APSPA4 0003-7028 Google Scholar
X. Min et al.,
“Double-beam near-infrared spectroscopy to correct light source drift in aqueous glucose solution experiments,”
Anal. Methods, 6
(24), 9831
–9840
(2014). http://dx.doi.org/10.1039/C4AY02178A AMNEGX 1759-9679 Google Scholar
N. Teramae and J. D. Winefordner,
“Double-beam lens spectroscopy with dual-beam configuration based on photo-differential detection,”
Appl. Spectrosc., 41
(1), 164
–165
(1987). http://dx.doi.org/10.1366/0003702874867945 APSPA4 0003-7028 Google Scholar
R. N. Johnson and J. R. Baker,
“Accuracy of devices used for self-monitoring of blood glucose,”
Ann. Clin. Biochem.: Int. J. Biochem. Lab. Med., 35
(1), 68
–74
(1998). http://dx.doi.org/10.1177/000456329803500108 Google Scholar
Y. Luo et al.,
“Discussion on floating-reference method for noninvasive measurement of blood glucose with near-infrared spectroscopy,”
Proc. SPIE, 6094 60940K
(2006). http://dx.doi.org/10.1117/12.645684 PSISDG 0277-786X Google Scholar
X. Guo, A. Mandelis and B. Zinman,
“Noninvasive glucose detection in human skin using wavelength modulated differential laser photothermal radiometry,”
Biomed. Opt. Express, 3
(11), 3012
–3021
(2012). http://dx.doi.org/10.1364/BOE.3.003012 BOEICL 2156-7085 Google Scholar
W. Zhang et al.,
“Discussion on the validity of NIR spectral data in non-invasive blood glucose sensing,”
Biomed. Opt. Express, 4
(6), 789
–802
(2013). http://dx.doi.org/10.1364/BOE.4.000789 BOEICL 2156-7085 Google Scholar
J. Liu, R. Liu and K. Xu,
“Accuracy of noninvasive glucose sensing based on near-infrared spectroscopy,”
Appl. Spectrosc., 69
(11), 1313
–1318
(2015). http://dx.doi.org/10.1366/14-07728 APSPA4 0003-7028 Google Scholar
W. Zhang et al.,
“Net analyte signal with floating reference theory in non-invasive blood glucose sensing by near-infrared spectroscopy,”
Chin. Opt. Lett., 10
(8), 083002
(2012). http://dx.doi.org/10.3788/COL CJOEE3 1671-7694 Google Scholar
X. Min et al.,
“A preliminary verification of the floating reference measurement method for non-invasive blood glucose sensing,”
Opt. Laser Technol., 91 7
–12
(2017). http://dx.doi.org/10.1016/j.optlastec.2016.12.011 Google Scholar
Y. Yang et al.,
“Reference point of floating-reference method for blood glucose sensing,”
Chin. Opt. Lett., 8
(4), 421
–424
(2010). http://dx.doi.org/10.3788/COL CJOEE3 1671-7694 Google Scholar
A. P. Popov,
“Effect of glucose concentration in a model light-scattering suspension on propagation of ultrashort laser pulses,”
Quantum Electron., 35
(35), 1075
–1078
(2005). http://dx.doi.org/10.1070/QE2005v035n11ABEH012794 QUELEZ 1063-7818 Google Scholar
A. P. Popov et al.,
“Glucose sensing in flowing blood and intralipid by laser pulse time-of-flight and optical coherence tomography techniques,”
IEEE J. Sel. Top. Quantum Electron., 18
(4), 1335
–1342
(2012). http://dx.doi.org/10.1109/JSTQE.2011.2175202 IJSQEN 1077-260X Google Scholar
J. B. Fishkin and E. Gratton,
“Propagation of photon-density waves in strongly scattering media containing an absorbing semi-infinite plane bounded by a straight edge,”
J. Opt. Soc. Am. A, 10
(1), 127
–140
(1993). http://dx.doi.org/10.1364/JOSAA.10.000127 Google Scholar
L. Wang, S. L. Jacques and L. Zheng,
“MCML—Monte Carlo modeling of light transport in multi-layered tissues,”
Comput. Methods Programs Biomed., 47
(2), 131
–146
(1995). http://dx.doi.org/10.1016/0169-2607(95)01640-F CMPBEK 0169-2607 Google Scholar
T. L. Troy and S. N. Thennadil,
“Optical properties of human skin in the near infrared wavelength range of 1000 to 2200 nm,”
J. Biomed. Opt., 6
(2), 167
–176
(2001). http://dx.doi.org/10.1117/1.1344191 JBOPFO 1083-3668 Google Scholar
M. Kohl, M. Cope and M. Essenpreis,
“The influence of glucose concentration upon the transport of light in tissue-simulating phantoms,”
Phys. Med. Biol., 40
(7), 1267
–1287
(1995). http://dx.doi.org/10.1088/0031-9155/40/7/009 Google Scholar
M. Kohl et al.,
“Influence of glucose concentration on light scattering in tissue-simulating phantoms,”
Opt. Lett., 19
(24), 2170
–2172
(1994). http://dx.doi.org/10.1364/OL.19.002170 OPLEDP 0146-9592 Google Scholar
Biography |