[Application of EMD algorithm to the dynamic spectrum non-invasive measurement of hemoglobin].

Empirical mode decomposition (EMD) algorithm combined with the theory of dynamic spectrum extraction at frequency domain was applied to the noninvasive measurement of hemoglobin concentration. Fifty seven cases' photoplethysmography was collected in the range of 636.98-1086.86 nm in vivo. After the denoising preprocess through the EMD method for each wavelength pulse wave of each sample separately, dynamic spectrum of each sample was made up of all peaks extracted by Fourier transform. Partial least squares regression model was used to establish the calibration and prediction of hemoglobin concentration. Compared to the modeling results without EMD, the correlation coefficient of predicted values and the real values was increased from 0.8798 up to 0.9176. The root mean square error of prediction set was reduced from 6.6759 to 5.3001 g x L(-1) and the relative error was reduced from 8.45 to 6.71%. The modeling accuracy has been greatly improved. The results showed that EMD algorithm can be effectively applied to denoise the spectral data and improve the accuracy of the non-invasive measurement of blood components.

[Noninvasive measurement of the human RBC concentration based on BP NN model].

In the present article, the BP neural network's arithmetic model is applied to the noninvasive detection of the concentration of the red blood cell of human body. Due to the use of BP neural network in the modeling and analysis of the dynamic spectrum data and the actual measured value of the red blood cell, the authors get a better result which refers to that the output value tracks the expected result very well. The related coefficient R can reach 0.993. When predicting the output value in the way of the BP neural network model, the maximal relative error is only 4.7%, average relative error is 2.1%, so the authors can say that it has more ideal prediction ability. The experimental result shows that the BP neural network model can be accurate in dealing with the nonlinear relation between the dynamic spectrum data and human erythrocyte practical value and it can make the method of noninvasive blood analysis more useful in clinical application. So it has a high application value.

[Noninvasive measurement of serum total protein content by near-infrared reflection spectra with tongue inspection].

The technology of tongue near-infrared reflectance spectra was used for human serum total protein (TP) content of noninvasive testing for the first time. Reflectance spectrum on the tongue tips of 58 volunteers was collected, and the biochemical values of serum total protein were recorded at the same time. The samples were separated into two parts: training set and prediction set. Two prediction models were established using PCA combined with BP neural network and PLS. Using PCA-BP model to predict the prediction set, the average relative error is 7.35%, RMSEP was 6.3771 g x L(-1), and the correlation coefficient was 0.9021. Using PLS model to predict the prediction set, the average relative error is 4.77%, RMSEP was 0.1304 g x L(-1), and the correlation coefficient was 0.9718. It was approved that reflectance spectra of tongue can be used to predict TP accurately and noninvasively.

[Improving component analysis ability of the complex mixed solutions by multi-dimensional diffuse transmittance spectrometry].

The multi-dimensional diffuse transmittance spectrums were collected by the traditional near-infrared transmittance method combined with a scanning device, and then used for component analysis of the complex mixed solution. A xenon light, an electric control translation stage and a spectrometer were gathered to set up a device; Intralipid-20%, India-ink and C6H12O6 were used to prepare 225 kinds of complex mixed solutions; the diffuse transmittance spectrums were measured at 20 points off the transmission center distributed from 0-5 mm (interval 0.25 mm); the single and multi-point diffuse transmittance spectrums were analyzed by partial least squares regression for modeling and prediction. The results show that the modeling and prediction accuracy of the concentrations of the intralipid-20% and India-ink increased with the growing of the transmittance points, but the concentration of the C6H12O6 did not increase. It is proved that the spectrums collected by different points can raise the signal to noise radio of the strong absorption and scattering substance, and the signal to noise radio of the weak absorption and scattering substance would be improved by increasing the current system accuracy.

[The effects of signal to noise ratio of instrument and number of wavelengths on the accuracy of spectral analysis].

In order to study the influence of the two factors, the signal to noise ratio (SNR) of instrument and number of wavelengths, on the accuracy of quantitative analysis in spectral modeling analysis, the influences of different SNRs, the number of wavelengths participating in partial least square regression, and multi-wavelength modeling in different SNR bands were taken into account. The relationships between these factors and quantitative analysis accuracy were explored by modeling analysis experiments. The results show that the analysis accuracy is directly related to the SNR of instrument. And increasing wavelengths, especial the ones with high SNR, reflects a phenomenon that four times of wavelength number will bring 2 times accuracy increasing, which can compensate for the lack of SNR. The research provides the experimental basis and theoretical guidance to improve the model's quality and reduce the prediction error by the utilization and improvement of the instrument's SNR and the reasonable choice of wavelengths and bands of modeling.

[The analysis for improving the SNR of blood components noninvasive measurement with DS method].

In order to increase the accuracy of blood components measurement and enhance the stability of prediction model, the quantitative signal-noise-ratio (SNR) analysis of measuring instruments based on dynamic spectrum (DS) and preprocessing method was conducted. The SNR of DS is increased after adding boxcar integrator, decreasing wavelength revolution, balancing the DS's SNR and excluding gross errors in preprocessing according to experiment results. Two volunteers were tested continuously for many times using the DS data acquiring system. The correlation coefficients of the each volunteer's DS data was increased from 0.934 and 0.953 to 0.991 and 0.987, respectively. Moreover, the gap between the correlation coefficient of the same volunteer's DS and different volunteers' DS is increased too, which shows that the SNR can be improved by these methods. The quantitative SNR analysis can guide the way of choosing preprocessing method efficiently, which will create the condition for clinical application of the blood components noninvasive measurement.

[Effect of different distribution of components concentration on the accuracy of quantitative spectral analysis].

In order to discuss the effect of different distribution of components concentration on the accuracy of quantitative spectral analysis, according to the Lambert-Beer law, ideal absorption spectra of samples with three components were established. Gaussian noise was added to the spectra. Correction and prediction models were built by partial least squares regression to reflect the unequal modeling and prediction results between different distributions of components. Results show that, in the case of pure linear absorption, the accuracy of model is related to the distribution of components concentration. Not only to the component we focus on, but also to the non-tested components, the larger covered and more uniform distribution is a significant point of calibration set samples to establish a universal model and provide a satisfactory accuracy. This research supplies a theoretic guidance for reasonable choice of samples with suitable concentration distribution, which enhances the quality of model and reduces the prediction error of the predict set.

[A modulation method to improve the signal-to-noise ratio of internal virtual hyperspectrum of body surface].

Virtual internal hyperspectrum of body surface which can express and properly mix the information of skin's composition and structure was expected to apply in the human disease quick-diagnosis, but the weak detected signal was easily impacted by ambient light and dark current from the instrument, so square wave modulated by chopper disk was proposed to resolve this issue. First, we built the experiment platform, and selected both laser and bromine tungsten lamp as the light source, then compared the virtual hyperspectrum of non-modulated and modulated light on human arm in the environment with artificial interference, and demodulated the modulated signal by FFT method. The experiment result showed that modulation method could effectively eliminate the impact brought by environmental light and dark current and improve the signal to noise ratio (SNR), suggesting a new approach to detect human's weak optical signals.

[The concentration resolution of complex liquid spectrometry analysis and the effect of concentration distribution].

In order to ensure the feasibility of complex liquid spectroscopy analysis, to analyze the accuracy gain of modeling by multi-wavelength, and to determine the appropriate distribution of concentration to obtain the high quality and universal quantitative analysis model, the precision of the detection of composition concentration by spectral analysis is illustrated through a error analysis which takes into account the following three contributions: spectral instrument noise, multi-wavelength modeling and the distribution of composition concentration. By concentration resolution analysis, the concentration resolution can be achieved when the spectrometer noise is available, but also the theoretical basis is provided to select a suitable spectrometer to meet the resolution requirement of quantitative analysis. Over-sampling technique indicates that the precision improvement by modeling with multi-wavelength can obtain higher concentration detection sensitivity. The sparse-dense-ratio and Euclidean distance of both measured and non-measured components provide the theoretic guidance for choosing the suitable concentration distribution which improves the model's quality and reduces the prediction error of the sample set.

[Study on detection of changes in tissue structure and composition by using virtual internal hyper-spectrum of body surface].

Purposed to achieve the detection of changes in tissue structure and composition simultaneously by virtual internal hyper-spectrum of body surface (VIHBS), we designed the VIHBS system containing modulated light, translation stage device for data acquisition and spectrometer. In the present study, pork meat as the experimental subject was equally divided into three pieces. One piece was used for a control study, and the others were separately embedded with red filter and injected into highly scattering intralipid to imitate the changes in tissue structure and composition. And then, data acquisition of 17 points started at source fiber was taken at intervals of 0.5 mm. The results showed that modulated light can effectively inhibit the influences of ambient light and dark current created by optoelectronic devices. In addition, VIHBS technology achieved the locating of red filter by visible-light and the rapid screening of changes in tissue composition caused by intralipid with near infrared light. The study suggests that it is much feasible to detect the changes in tissue structure and composition by VIHBS. And it is concluded that further research will be likely to provide a new method to realize the early diagnosis and screening of human skin.

[Feasibility of composition analysis of complex mixed solution by hyperspectral technique].

The complex mixed solution is a common material form in all walks of life. It was difficult to achieve the desired results if the complex mixed solution was detected by the traditional spectral analysis method. The hyperspectral technology was taken to detect the complex mixed solution to improve the signal to noise ratio of the spectrum by utilizing the absorbing and scattering characteristics of the analytes at the same time. A hyperspectral acquisition device was designed to collect the diffuse reflectance hyperspectral images of the analytes (Intralipid-10%). The Monte Carlo simulation and the diffuse approximation were used to validate the experimental device. The authors found that the absorption coefficient of the Intralipid-10% at 632 nm was 0.002 0 cm(-1) and the reduced scattering coefficient was 63.35 cm(-1); the corresponding relative error of the standard reference was 11.1% and 6.49%. The inversion result of the diffuse approximation validated the exactness of the experimental device. Finally, the hyperspectral images of milk and fruit juice from different manufacturers were taken, the images show that the differences between different samples were more obvious than that of traditional 2-dimensional spectrum. This research reveals that the hyperspectral technology is feasible in the component analysis of complex mixed solution.

[Equalization of whole-band signal's SNR in the blood components noninvasive measurement].

To fully extend the category of blood components that can be noninvasively measured by dynamic spectrum (DS) method and to increase its measuring precision, an overall consideration of light source, tissue absorption and sensor's sensitivity was made. Compensating the light source and adding the telecentric lens not only expand the spectral effective detecting range, but also balance the signal-to-noise ratio (SNR) of the photoelectric pulse in the whole band equalization. The integral SNR of DS signal was increased and the measurement bandwidth was expanded. The effectiveness of this method was validated by the quality evaluation criterion of DS data: the effective detecting range of visible DS was widened from 600-1 000 nm to 500-1 135 nm; the effective detecting range of near-infrared DS was widened from 900-1 100 nm to 900-1 700 nm. The results show that the design can create the condition for detection of new blood components noninvasively, and enhance the prediction accuracy of the blood components, for which noninvasive measuring using DS method has been achieved.

[Component analysis of complex mixed solution based on multidimensional diffuse reflectance spectroscopy].

In the present paper, the authors proposed a method for component analysis of complex mixed solutions based on multidimensional diffuse reflectance spectroscopy by analyzing the information carried by spectrum signals from various optical properties of various components of the analyte. The experiment instrument was designed with supercontinuum laser source, the motorized precision translation stage and the spectrometer. The Intralipid-20% was taken as an analyte, and was diluted over a range of 1%-20% in distilled water. The diffuse reflectance spectrum signal was measured at 24 points within the distance of 1.5-13 mm (at an interval of 0.5 mm) above the incidence point. The partial least squares algorithm model was used to perform a modeling and forecasting analysis for the spectral analysis data collected from single-point and multi-point. The results showed that the most accurate calibration model was created by the spectral data acquired from the nearest 1-13 points above the incident point; the most accurate prediction model was created by the spectral signal acquired from the nearest 1-7 points above the incident point. It was proved that multidimensional diffuse reflectance spectroscopy can improve the spectral signal to noise ratio. Compared with the traditional spectrum technology using a single optical property such as absorbance or reflectance, this method increased the impact of scattering characteristics of the analyte. So the use of a variety of optical properties of the analytes can make an improvement of the accuracy of the modeling and forecasting, and also provide a basis for component analysis of the complex mixed solution based on multidimensional diffuse reflectance spectroscopy.

Relaxation creep rupture of heterogeneous material under constant strain.

We focus on a system consisting of an elastic part and a damageable part in series, to study the relaxation creep rupture of a heterogeneous system subjected to a uniaxial constant strain applied instantaneously. The viscoelastic behavior of the damageable part is modeled by a fiber bundle model consisting of Kelvin-Voigt elements and global load sharing is assumed for the redistribution of load following fiber breaking in the damageable part. Analytical and numerical calculations show that the global relaxation creep rupture appears if the elastic energy stored in the elastic part exceeded the fracture energy of the damageable part. The lifetime of the system strongly depends on the values of the applied external strain and the initial stiffness ratio k between the elastic part and the damageable part. We show that a higher stiffness ratio implies a more brittle system. Prior to complete failure, relaxation creep rupture exhibits a sequence of three stages, similar to creep rupture under constant stress, and the nominal force rate presents a power law singularity with a power index -1/2 near the global rupture time.

[D-value estimation of dynamic spectrum based on the statistical methods].

To realize the noninvasive concentration detection of blood components and improve the drawbacks existing in the time-domain single-trial estimation method of "dynamic spectrum" (DS), the D-value estimation method based on the statistical properties was proposed. We extracted the absolute difference between two corresponding values of each wavelength to make up the DS, selected the valid DSs from the DSs of different times by statistic method, and the valid DSs were superimposed and averaged as the final output of the DS. Data collected from 48 volunteers were processed by the D-value estimation and the single-trial estimation, respectively; and then the comparison was carried out between the two methods. Compared with the single-trial estimation, the valid DSs extracted by the D-value estimation were slightly better in denoising; And the average number of the remained valid DSs is improved from 48 to 130; the average of mean square error among the valid DSs is improved from 0.39 to 0.006; the speed of data processing is increased by nearly 20 times. The new method can significantly improve the quality of the extraction of DS.

[Application of EMD and SPA algorithm in the detection of benzoyl peroxide addition in flour by spectroscopy].

In order to improve the accuracy of spectroscopy analysis and reduce the modeling wavelength numbers, empirical mode decomposition (EMD) and successive projections algorithm (SPA) were applied together in the measurement of benzoyl peroxide (BPO) additive amount in flour by near infrared spectroscopy. Spectra of flour samples into which BPO were added were collected. Firstly, EMD was implemented to eliminate the noise of original spectra, and then SPA was employed to select the characteristic wavelengths. The precision of the model based on the processed spectra by EMD was greatly improved compared with the model based on the original spectra, with the calibration determination coefficient Rcal2 increased from 0.81 to 0.899 and the prediction determination coefficient R2pred increased from 0. 755 4 to 0. 86. Seven characteristic variables were selected from 512 wavelength variables by SPA. And the performance of the model built by the selected characteristic variables (Rcal2 is 0.863, Rpred2 is 0.86) was as good as full-spectrum model's, while the number of modeling variables was greatly reduced by 96.4%. The results indicated that empirical mode decomposition and successive projections algorithm can be effectively used to denoise the spectra and selected characteristic wavelengths for the detection of BPO addition in flour. The seven selected wavelengths in this paper can be a reference for designing portable BPO detection meter.

[Research on the measurement of urinary albumin by visible-near infrared spectroscopy].

The urinary albumin (UMALB) is the most reliable diagnostic indicator of renal injury in clinical. Attempting to realize the rapid and free reagent measurement of UMALB, the visible-near infrared multiple optical path length spectra of 207 urine samples were collected. By the nonlinear characteristics of multiple optical path length spectra, more information about the component of sample contents can be obtained. The PLS model of the spectra and UMALB was firstly established. Based on it, the PLS-ANN modeling method was built to introduce nonlinear information. By contrast, the PLS-ANN modeling method can obtain a better model to improve the accuracy of quantitative analysis. The R2 of predicted model was 0.951 1 and the RMSEP was 5.02 mg x L(-1). The results showed the feasibility of the visible-near infrared multiple optical path length spectroscopy technique for urinary albumin analysis. This research establishes the foundation of detecting the urinary albumin and other components free of reagent conveniently and rapidly.

[Single-trial estimation of dynamic spectrum].

To improve the efficiency and accuracy of the dynamic spectrum data processing, the method of single-trial estimation was adopted. First, the rising edge of the whole band PPG was extracted, which was calculated by averaging superimposed collected photoelectric plethysmography (PPG) at all wavelengths as a template per single pulse; Second, this template was used to correct the rising edge of PPG at all wavelengths, and the difference of absorbance was calculate, and then a single-trial DS was obtained; finally, the single-trial DS which contained the gross error under the 3sigma criterion was removed, and then the remaining superimposed single-trial DS was averaged as the final output of the DS. Data measured from 10 volunteers were compared with the results of the extraction in frequency domain: the correlation coefficient distribution of the DS from the same finger of the same individual was improved from 0.006775 to 0.0003840; the correlation coefficient distribution of the DS from the different fingers of the same individual was improved from 0.01393 to 0.002205, whereas the differences of DS between different indivisuals with high quality, and accelerate the process of DS put into practical application.

[Noninvasive prediction of red blood cell counts by normalized reflection spectroscopy for tongue inspection].

The technology of visible-infrared (Vis-NIR) spectroscopy for tongue inspection was used for the noninvasive prediction of red blood cell counts (RBC). Reflectance spectra on the tongue tips of 240 volunteers were collected, and the tongue picture was recorded at the same time. Samples were separated into two parts: calibration sample and prediction sample. Spectra were then subjected to a partial least squares regression (PLSR) analysis. The correlation between the data and prediction of HCT yielded calibration samples value of 0.991, and RBC levels of prediction samples predicted by this model from Vis-NIR spectra provided a coefficient of determination in prediction of 0.994 with an average relative error of prediction of 0.81%, and a root mean square error of prediction of 0.472. Experiment results showed that the correlation between tongue and RBC exists, and also demonstrated that the tongue can be used as the measure site for noninvasive prediction of RBC. From the results of prediction model and prediction results we can conclude that reflectance spectroscopy of tongue can be used to predict the RBC noninvasively.

[Rapid diagnosis of TCM syndrome based on spectrometry].

Tongue color provides important information in traditional Chinese medicine (TCM) diagnosis, but in the process of TCM inspection, the various surrounding environment and subjective effect of doctors will influence the correctness of diagnosis. The present article put forward a brand new thought to study TCM objectivity, that is to research the essence message of tongue from perspective of spectrum, study the continuous spectrum to replace the observation of tongue color. The experiment used near infrared spectrum to collect reflection spectrum data of tongue among 53 exterior cold and interior heat patients, 37 healthy people and 21 wind chill suffers. Used matlab for data pretreatment, minitab for statistics modeling and prediction used partial least squares, the accuracy of prediction is 85.6%, but however, the spectrum of near infrared fail to distinguish exterior cold and interior heat patients from wind chill sufferers. The article provides a brand new way to implement the objectivity of tongue diagnosis in TCM, and also offers data support for the study of TCM syndrome.