Calibration in non-linear NIR spectroscopy using principal component artificial neural networks

Near-infrared (NIR) spectroscopy was used in simultaneous, non-destructive analysis of antipyriine and caffeine citrate tablets. Principal component artificial neural networks (PC-ANNs) were used to construct models for the analytes, using the testing set for external validation. Four pretreated spectra, namely, first-derivative, second-derivative, standard normal variate (SNV) and multiplicative scatter correction (MSC) spectra led to simplified and more robust models than conventional spectra. In PC-ANNs models, the spectra data were analyzed by principal component analysis (PCA) firstly. Then the scores of the principal compounds (PCs) were chosen as input nodes for input layer instead of the spectra data. The artificial neural networks (ANNs) models using the spectra data as input nodes were also established, which were compared with the PC-ANNs models. The result shows the SNV model of PC-ANNs multivariate calibration has the lowest training error and predicting error. The concept of the degree of approximation was introduced and performed as the selective criterion of the optimum network parameters.     

Artificial neural network for simultaneous determination of two components of compound paracetamol and diphenhydramine hydrochloride powder on NIR spectroscopy

Diffuse reflectance near-infrared (NIR) spectroscopy is a technique widely used for rapid and non-destructive analysis of solid samples. A method for simultaneous analysis of the two components of compound paracetamol and diphenhydramine hydrochloride powdered drug has been developed by using artificial neural network (ANN) on near-infrared (NIR) spectroscopy. An ANN containing three layers of nodes was trained. Various ANN models based on pretreated spectra (first-derivative, second-derivative and standard normal variate; SNV) were tested and compared, respectively. In the models the concentration of paracetamol and caffeine as active principles of compound paracetamol and diphenhydramine hydrochloride powder was determined simultaneously. Partial least squares regression (PLS) multivariate calibrations were also used, which were compared with ANN. The best model was obtained at first-derivative spectra. We have also discussed the parameters that affected the networks and predicted the test set (unknown) specimens. The degree of approximation, a new evaluation criteria of the network were employed, which proved the accuracy of the predicted results.     

Radial basis function neural networks in non-destructive determination of compound aspirin tablets on NIR spectroscopy

The application of the second most popular artificial neural networks (ANNs), namely, the radial basis function (RBF) networks, has been developed for quantitative analysis of drugs during the last decade. In this paper, the two components (aspirin and phenacetin) were simultaneously determined in compound aspirin tablets by using near-infrared (NIR) spectroscopy and RBF networks. The total database was randomly divided into a training set (50) and a testing set (17). Different preprocessing methods (standard normal variate (SNV), multiplicative scatter correction (MSC), first-derivative and second-derivative) were applied to two sets of NIR spectra of compound aspirin tablets with different concentrations of two active components and compared each other. After that, the performance of RBF learning algorithm adopted the nearest neighbor clustering algorithm (NNCA) and the criterion for selection used a cross-validation technique. Results show that using RBF networks to quantificationally analyze tablets is reliable, and the best RBF model was obtained by first-derivative spectra.     

Qualitative and quantitative analysis of oxytetracycline by near-infrared spectroscopy

Near-infrared (NIR) spectroscopy, in combination with chemometrics, enable the analysis of raw materials without time-consuming sample preparation methods. The aim of our work was to estimate critical parameters in the analytical specification of oxytetracycline, and consequently the development of a method for quantification and qualification of these parameters by NIR spectroscopy. A Karl Fischer (K.F.) titration to determine the water content, a colorimetric assay method, and Fourier transform-infrared (FT-IR) spectroscopy to identify the oxytetracycline base, were used as reference methods, respectively. Multivariate calibration was performed on NIR spectral data using principal component analysis (PCA), partial least-squares (PLS 1) and principal component regression (PCR) chemometric methods. Multivariate calibration models for NIR spectroscopy have been developed. Using PCA and the Soft Independent Modelling of Class Analogy (SIMCA) approach, we established the cluster model for the determination of sample identity. PLS 1 and PCR regression methods were applied to develop the calibration models for the determination of water content and the assay of the oxytetracycline base. Comparing the PLS and PCR regression methods we found out that the PLS is better established by NIR, especially as the spectroscopic data (NIR spectra) are highly collinear and there are many wavelengths due to non-selective wavelengths. The calibration models for NIR spectroscopy are convenient alternatives to the colorimetric method and to the K.F. method, as well as to FT-IR spectroscopy, in the routine control of incoming material.     

应用光谱技术无损检测油菜叶片中乙酰乳酸合成酶

应用可见/近红外光谱技术实现了油菜叶片中乙酰乳酸合成酶(ALS)的快速无损检测.对99个油菜样本进行光谱扫描,经过平滑、变量标准化、一阶求导等预处理后,应用偏最小二乘法(PLS)建立了ALS的预测模型.同时提取有效特征变量,作为反向传输人工神经网络(BPNN)和最小二乘-支持向量机(LS-SVM)的输入值,并建立相应的模型.用66个样本建模,33个样本验证.结果表明,LS-SVM模型能够获得最优的预测结果,预测集样本的相关系数(r)、预测标准差(RMSEP)和偏差(Bias)分别为0.998、 0.715和0.079,获得了满意的预测精度.结果表明,应用可见/近红外光谱技术结合LS-SVM检测油菜中乙酰乳酸合成酶是可行的,并能获得满意的预测精度,为进一步应用光谱技术进行油菜生长状况的大田监测奠定了基础.     

Non-destructive determination of metronidazole powder by using artificial neural networks on short-wavelength NIR spectroscopy

The present study aimed at providing a new method in sight into short-wavelength near-infrared (NIR) spectroscopy of in pharmaceutical quantitative analysis. To do that, 124 experimental samples of metronidazole powder were analyzed using artificial neural networks (ANNs) in the 780-1100 nm region of short-wavelength NIR spectra. In this paper, metronidazole was as active component and other two components (magnesium stearate and starch) were as excipients. Different preprocessing spectral data (first-derivative, second-derivative, standard normal variate (SNV) and multiplicative scatter correction (MSC)) were applied to establish the ANNs models of metronidazole powder. The degree of approximation, a new evaluation criterion of the networks was employed to prove the accuracy of the predicted results. The results presented here demonstrate that the short-wavelength NIR region is promising for the fast and reliable determination of major component in pharmaceutical analysis.     

Application of Near Infrared Diffuse Reflectance Spectroscopy with Radial Basis Function Neural Network to Determination of Rifampincin Isoniazid and Pyrazinamide Tablets

Partial least squares(PLS),back-propagation neural network(BPNN)and radial basis function neural network(RBFNN)were respectively used for estalishing quantative analysis models with near infrared(NIR)diffuse reflectance spectra for determining the contents of rifampincin(RMP),isoniazid(INH)and pyrazinamide(PZA)in rifampicin isoniazid and pyrazinamide tablets.Savitzky-Golay smoothing,first derivative,second derivative,fast Fourier transform(FFT)and standard normal variate(SNV)transformation methods were applied to pretreating raw NIR diffuse reflectance spectra.The raw and pretreated spectra were divided into several regions,depending on the average spectrum and RSD spectrum.Principal component analysis(PCA)method was used for analyzing the raw and pretreated spectra in different regions in order to reduce the dimensions of input data.The optimum spectral regions and the models' parameters were chosen by comparing the root mean square error of cross-validation(RMSECV)values which were obtained by leave-one-out cross-validation method.The RMSECV values of the RBFNN models for determining the contents of RMP,INH and PZA were 0.00288,0.00226 and 0.00341,respectively.Using these models for predicting the contents of INH,RMP and PZA in prediction set,the RMSEP values were 0.00266,0.00227 and 0.00411,respectively.These results are better than those obtained from PLS models and BPNN models.With additional advantages of fast calculation speed and less dependence on the initial conditions,RBFNN is a suitable tool to model complex systems.     

Orthogonal projection to latent structures combined with artificial neural network for quantitative analysis of phenoxymethylpenicillin potassium powder

A method for quantitative analysis of phenoxymethylpenicillin potassium powder on the basis of near-infrared (NIR) spectroscopy is investigated by using orthogonal projection to latent structures (O-PLS) combined with artificial neural network (ANN). Being a preprocessing method, O-PLS can remove systematic orthogonal variation from a given data set X without disturbing the correlation between X and the response set y. In this paper, O-PLS method was applied to preprocess the original spectral data of phenoxymethylpenicillin potassium powder, and the filtered data was used to establish the ANN model. In this model, the concentration of phenoxymethylpenicillin potassium as the active component was determined. The degree of approximation was employed as the selective criterion of the optimum network parameters. In order to compare with O-PLS-ANN model, the calibration models that use the original spectra and different preprocessing methods (first-derivative, second-derivative, standard normal variate (SNV) and multiplicative scatter correction (MSC)) of the spectra were also designed. Experimental results show that O-PLS-ANN model is the best.     

Enhancing calibration models for non-invasive near-infrared spectroscopical blood glucose determination

Partial least-squares regression (PLS) and radial basis function (RBF) networks are used to compute calibration models for non-invasive blood glucose determination by NIR diffuse reflectance spectroscopy. A model computation shows that even extremely small deviations of the spectra induce increased prediction errors. Since the spectral contribution of blood glucose is much smaller than deviations resulting from the non-invasive measuring process a method based on Pearson’s correlation coefficient can be used for evaluating the quality of the recorded spectra during the prediction step. Another method is based on the leverage values from the hat matrix of the RBF network. Both methods lead to a significant decrease in prediction error.     

Determination of acetolactate synthase activity and protein content of oilseed rape (Brassica napus L.) leaves using visible/near-infrared spectroscopy

A new acetolactate synthase (ALS)-inhibiting herbicide, propyl 4-(2-(4,6-dimethoxypyrimidin-2-yloxy)benzylamino)benzoate (ZJ0273), was applied to oilseed rape (Brassica napus L.) leaves in different leaf positions. Visible/near-infrared (Vis/NIR) spectroscopy was investigated for fast and non-destructive determination of ALS activity and protein content in rapeseed leaves. Partial least squares (PLS) analysis was the calibration method with comparison of different spectral preprocessing by Savitzky-Golay (SG) smoothing, standard normal variate (SNV), first and second derivative. The best PLS models were obtained by first-derivative spectra for ALS, whereas original spectra for soluble, non-soluble and total protein contents. Simultaneously, certain latent variables (LVs) were used as the inputs of back-propagation neural network (BPNN) and least squares-support vector machine (LS-SVM) models. All LS-SVM models outperformed PLS models and BPNN models. The correlation coefficient (r), root mean square error of prediction (RMSEP) and bias in validation set by LS-SVM were 0.998, 0.715 and 0.079 for ALS, 0.999, 33.084 and 1.178 for soluble protein, 0.997, 42.773 and 6.244 for non-soluble protein, 0.999, 59.562 and 7.437 for total protein, respectively. The results indicated that Vis/NIR spectroscopy combined with LS-SVM could be successfully applied for the determination of ALS activity and protein content of rapeseed leaves. The results would be helpful for further on field analysis of using Vis/NIR spectroscopy to monitor the growing status and physiological properties of oilseed rape.     

Butene concentration prediction in ethylene/propylene/1-butene terpolymers by FT-IR spectroscopy through multivariate statistical analysis and artificial neural networks

This paper reports the development of calibration models for quality control in the production of ethylene/propylene/1-butene terpolymers by the use of multivariate tools and FT-IR spectroscopy.1-Butene concentration prediction is achieved in terpolymers by coupling FT-IR spectroscopy to multivariate regression tools. A dataset of 26 terpolymers (14 coming from a constrained experimental design for mixtures, plus 12 terpolymers used for external validation) was analysed by FT-IR spectroscopy. An internal method of “Polimeri Europa” plant, based on ¹³C NMR spectroscopy is used to determine the percentage of 1-butene in the samples. Then, different multivariate tools are used for 1-butene concentration prediction based on the FT-IR spectra recorded. Different multivariate calibration methods were explored: principal component regression (PCR), partial least squares (PLS), stepwise OLS regression (SWR) and artificial neural networks (ANNs). The model obtained by back-propagation neural networks turned out to be the best one. The performances of the BP-ANN model were further improved by variable selection procedures based on the calculation of the first derivative of the network.The proposed approach allows the monitoring in real time of the polymer synthesis and the estimation of the characteristics of the product attainable from the concentration of 1-butene.     

Optical coefficient‐based multivariate calibration on near‐infrared spectroscopy

The time and expense of calibration development limit the feasibility of NIR spectroscopy for many industrial applications, with a major portion of the costs being related to creation of a sufficient set of calibration samples. Net analyte signal (NAS) and generalized least squares (GLS) pre‐processing have been proposed in the literature as methods to simplify multivariate calibration by reducing the quantity of calibration samples by orthogonalizing or shrinking interference signals. Synthetic calibration has also been reported as a method to combine interference signals with pure component spectra to generate virtual calibration models, thereby reducing the number of real calibration samples required. The goals of this paper were to (1) compare theoretical and practical differences between NAS and GLS pre‐processing and (2) explore the potential of simplified NIR calibrations, both empirical and synthetic, constructed using optical coefficient‐based signal processing on predicting chemical compositions of pharmaceutical powder mixtures. A reduced calibration dataset including only one pharmaceutical powder mixture composition and pure component spectra was used for both empirical and synthetic calibrations. Absorption and reduced scattering coefficients, obtained from spatially‐resolved spectroscopy, were used herein as interference signals in NAS/GLS pre‐processing for both calibrations. As a result, NAS and GLS were shown to be equivalent in both theoretical and practical senses. After optical coefficient‐based signal processing, simplified calibrations, both empirical and synthetic, were demonstrated to have similar model performance as generic pre‐processing methods such as SNV and derivative, while requiring fewer principal components and achieving a lower prediction error. Copyright © 2010 John Wiley & Sons, Ltd.     

Development and Application of Chemometric Methods in Near Infrared Spectral Analysis

In recent 10 years, like other disciplines influenced by the fast development of PC technique, chemometrics has been used in many analytical methods, especially in instrumental analysis. This article describes applications and comparison of multivariate linear regression (MLR), principal component analysis (PCA), principal component regression (PCR), partial least square (PLS), neural network (ANN), fuzzy and model recognition. A better calibration method can be a great help to improve the efficiency of the routine analytical work.     

Multicomponent kinetic spectrophotometric determination of pefloxacin and norfloxacin in pharmaceutical preparations and human plasma samples with the aid of chemometrics

A spectrophotometric method for the simultaneous determination of the important pharmaceuticals, pefloxacin and its structurally similar metabolite, norfloxacin, is described for the first time. The analysis is based on the monitoring of a kinetic spectrophotometric reaction of the two analytes with potassium permanganate as the oxidant. The measurement of the reaction process followed the absorbance decrease of potassium permanganate at 526nm, and the accompanying increase of the product, potassium manganate, at 608nm. It was essential to use multivariate calibrations to overcome severe spectral overlaps and similarities in reaction kinetics. Calibration curves for the individual analytes showed linear relationships over the concentration ranges of 1.0-11.5mgL(-1) at 526 and 608nm for pefloxacin, and 0.15-1.8mgL(-1) at 526 and 608nm for norfloxacin. Various multivariate calibration models were applied, at the two analytical wavelengths, for the simultaneous prediction of the two analytes including classical least squares (CLS), principal component regression (PCR), partial least squares (PLS), radial basis function-artificial neural network (RBF-ANN) and principal component-radial basis function-artificial neural network (PC-RBF-ANN). PLS and PC-RBF-ANN calibrations with the data collected at 526nm, were the preferred methods-%RPE(T) approximately 5, and LODs for pefloxacin and norfloxacin of 0.36 and 0.06mgL(-1), respectively. Then, the proposed method was applied successfully for the simultaneous determination of pefloxacin and norfloxacin present in pharmaceutical and human plasma samples. The results compared well with those from the alternative analysis by HPLC.     

Near-infrared spectra combined with partial least squares for pH determination of toothpaste of different brands

Near-infrared spectroscopy(NIR),which is generally used for online monitoring of the food analysis and production process, was applied to determine the internal quality of toothpaste samples.It is acknowledged that the spectra can be significantly influenced by non-linearities introduced by light scatter,therefore,four data preprocessing methods,including off-set correction, 1st-derivative,standard normal variate(SNV) and multiplicative scatter correction(MSC),were employed before the date analysis. The multivariate calibration model of partial least squares(PLS) was established and then was used to predict the pH values of the toothpaste samples of different brand.The results showed that the spectral date processed by MSC was the best one for predicting the pH value of the toothpaste samples.     

Combined wavelet transform-artificial neural network use in tablet active content determination by near-infrared spectroscopy

The pharmaceutical industry faces increasing regulatory pressure to optimize quality control. Content uniformity is a basic release test for solid dosage forms. To accelerate test throughput and comply with the Food and Drug Administration's process analytical technology initiative, attention is increasingly turning to nondestructive spectroscopic techniques, notably near-infrared (NIR) spectroscopy (NIRS). However, validation of NIRS using requisite linearity and standard error of prediction (SEP) criteria remains a challenge. This study applied wavelet transformation of the NIR spectra of a commercial tablet to build a model using conventional partial least squares (PLS) regression and an artificial neural network (ANN). Wavelet coefficients in the PLS and ANN models reduced SEP by up to 60% compared to PLS models using mathematical spectra pretreatment. ANN modeling yielded high-linearity calibration and a correlation coefficient exceeding 0.996.     

Near-infrared determination of active substance content in intact low-dosage tablets

Near-infrared (NIR) spectroscopy can be applied to determine the active substance content of tablets. Its great advantage lies in the minimal sample preparation required, which helps to reduce the potential for error. The aim of this study is to show the feasibility of this method on low-dosage tablets. The influence of various spectral pretreatments [standard normal variate (SNV), multiplicative scatter correction (MSC), second derivative (D2), orthogonal signal correction (OSC), separately and combined] and regression methods on prediction error are compared. Partial least square (PLS) regression provided better prediction than principal component regression (PCR). SNV was applied to the first data set and SNV and a second derivative to the second set to maximise model accuracy for quantifying the active substance of intact pharmaceutical products using diffuse reflectance NIR. The models yielded standard errors of prediction (SEP) of 0.1768 and 0.0682 mg for the two products. The experiments were conducted with two low-dosage pharmaceutical forms and results of NIR predictions were comparable to currently approved methods. Diffuse reflectance NIR has the potential to become a reliable and robust quality control method for determining active tablet content.     

Development of a rapid process monitoring method for dry-coated tableting process by using near-infrared spectroscopy

A nondestructive transmittance near-infrared (NIR) method for detecting off-centered cores in dry-coated (DC) tablets was developed as a monitoring system in the DC tableting process. Caffeine anhydrate was used as a core active pharmaceutical ingredient (API), and DC tablets were made by the direct compression method. NIR spectra were obtained from these intact DC tablets using the transmittance method. The reference assay was performed with HPLC. Calibration models were generated by partial least squares (PLS) regression and principal component regression (PCR) utilizing external validations. Hierarchical cluster analysis (HCA) of the results confirmed that NIR spectroscopy correctly detected off-centered cores in DC tablets. We formulated and used the Centering Index (CI) to evaluate the precision of core alignment and generated an NIR calibration model that could correctly predict this index. The principal component (PC) 1 loading vector of the final calibration model indicated that it could specifically detect the misalignment of tablet cores. The model also had good linearity and accuracy. The CIs of unknown sample tablets predicted by the final calibration model and those calculated through the HPLC analysis were closely parallel with each other. These results demonstrate the validity of the final calibration model and the utility of the transmittance NIR spectroscopic method developed in this study as a monitoring system in DC tableting process.     

Orthogonal projection to latent structures combined with artificial neural networks in non-destructive analysis of Ampicillin powder

A new method orthogonal projection to latent structures (O-PLS) combined with artificial neural networks is investigated for non-destructive determination of Ampicillin powder via near-infrared (NIR) spectroscopy. The modern NIR spectroscopy analysis technique is efficient, simple and non-destructive, which has been used in chemical analysis in diverse fields. Be a preprocessing method, O-PLS provides a way to remove systematic variation from an input data set X not correlated to the response set Y, and does not disturb the correlation between X and Y. In this paper, O-PLS pretreated spectral data was applied to establish the ANN model of Ampicillin powder, in this model, the concentration of Ampicillin as the active component was determined. The degree of approximation was employed as the selective criterion of the optimum network parameters. In order to compare the OPLS-ANN model, the calibration models that using first-derivative and second-derivative preprocessing spectra were also designed. Experimental results showed that the OPLS-ANN model was the best.     

Hydroxyl and acid number prediction in polyester resins by near infrared spectroscopy and artificial neural networks

Back-propagation artificial neural networks (BP-ANN) are applied for modeling hydroxyl number and acid value of a set of 62 samples of polyester resins from their near infrared (NIR) spectra. The results are compared to the classical calibration approaches, i.e. principal component regression (PCR) and partial least squares (PLS). The set of available samples is split into: (i) a training set, for models calculation; (ii) a test set, for setting the correct number of latent variables in PCR and PLS and for selecting the end point of the training phase of BP-ANN; (iii) a “production set” of samples, which are predicted to evaluate the models predictive ability. This approach guarantees that the predictive ability of the models is evaluated by genuine predictions. BP-ANN resulted always better than the classical PCR and PLS, from the point of view of the predictive ability. The study of the breakdown number of experiments to include in the training set showed instead that this factor does influence PCR and PLS at a lesser degree than what happens for BP-ANN. The latter approach requires a larger number of experiments for obtaining good results. The choice of optimal training sets is efficiently performed by Kohonen self-organizing maps (SOMs). It can be concluded that FT-NIR spectroscopy and BP-ANN models can be properly employed for monitoring the polyesterification of dicarboxylic acids with diols by predicting the acid and hydroxyl numbers directly along the process line.