Search

Raman spectroscopy represents an advanced process analytical technology to monitor and control chemical and biochemical processes. This study presents an autoencoder-based methodology that simulates Raman spectra from process variables and predicts the concentrations of different chemicals. The methodology accurately predicts concentrations from the spectra, even considering the temperature influences, and can work as an anomaly detector in process monitoring. The proposed methodology has significant implications for the optimization of industrial processes, improving process efficiency, reducing waste, and minimizing costs. It can also be extended to other industrial processes and imaging spectroscopy techniques, making it a valuable tool for process monitoring. This study highlights the effectiveness of autoencoders in simulating spectra and quantitative analysis, contributing significantly to the field of process monitoring. It has the potential to revolutionize industrial process monitoring and optimization, leading to substantial improvements in productivity and sustainability. [Display omitted] • Raman spectroscopy is a cutting-edge PAT for (bio)chemical process monitoring. • A method is introduced for simulating Raman spectra from process variables. • The method accurately predicts chemical concentrations with temperature variations. • The method can serve as an effective anomaly detector during process monitoring. [ABSTRACT FROM AUTHOR]

Salmon fillet was analysed via absorption spectroscopy with a hyperspectral camera (400–1000 nm) for four consecutive storage days. Optimised K-Nearest Neighbours analyses was performed on all variables (wavelengths) returning an average classification accuracy of 77.0 % for storage day prediction (across all days). A five principal component (PC) model returned an average accuracy of 73.7 % (all days). A histogram analysis of spectral pixels in the hyperspectral images matching reference spectra displayed increases in spoilt regions as the fillet aged for a Tail Bottom fillet section. Overall, five out of 12 of the fillet areas showed a monotonic increase in spoilage (p-value≈0.01). While Principal Component Analysis showed minimal separation in foremost PCs between days, this was improved by using 'fresh' and 'spoilt' class labels prescribed by the hyperspectral data. Via mean spectrum analysis, the dampening of an absorbance band around 600 nm was identified as the main discriminator between fresh and spoilt datasets. Therefore, spatial inhomogeneity of sample freshness should be considered in fish freshness studies, and hyperspectral imaging can be useful as tool to do so. [Display omitted] • Spectroscopy is emerging as a powerful tool to rapidly characterise the composition of different foods, which can be applied to food security. • Hyperspectral imaging, which combines spectroscopic and location information, can be used to ascertain spatial inhomogeneity in food samples. • Histogram analysis of images using a similarity algorithm isuseful to visualize spectroscopic differences. [ABSTRACT FROM AUTHOR]

Spectroscopy-based methods are often used to verify the authenticity of food and feed. In the scientific environment, especially non-targeted approaches are becoming increasingly popular here, although their routine applicability, e.g., in official food control, is currently limited. The comparability of spectra acquired by different spectroscopic instruments using the same measurement principle is essential for the application of a common spectral database in laboratories and represents a current challenge in this field. In order to investigate possible approaches towards improved comparability of these spectra, a sample set of rapeseed (Brassica napus) and pumpkin seed (Cucurbita maxima) oils was analyzed using three Fourier-transform-mid-infrared spectrometers following the same procedure for sample preparation and measuring conditions. Depending on the instrument, the obtained spectra exhibited differences in absorbance, but not in wavelengths of bands position and maxima. As a result, the mutual prediction of original data (without mathematical correction) from different instruments using a partial least squares discriminant analysis model built and optimized with spectral data from only one instrument provided a sensitivity of 0% for the two-class model of rapeseed in pumpkin seed oil, indicating no discrimination at all. In order to minimize the spectral differences and thus to enhance the performance parameters of the prediction models, different mathematical correction approaches were investigated: (i) instrument-specific correction factors, (ii) combinations of different pre-processing steps and (iii) piecewise direct standardization (partial least squares-based regression). All investigated approaches achieved classification results with 100% sensitivity for the mathematically corrected data set of the respective instrument. The outcome of this study indicates that the correction approaches can be used to optimize the comparability of spectral data from different instruments, which is a first step towards harmonization in non-targeted analysis. [Display omitted] • Comparison of MIR spectra acquired using three different instruments. • Different approaches to improve the comparability of spectra were tested. • A Calibration transfer approach was investigated. • Assessment of approaches by performance parameters of classification model. • Most promising results based on piecewise direct standardization. [ABSTRACT FROM AUTHOR]

Abstract Spectroscopic data generated by several PAT technologies is routinely used for the rapid assessment of quality properties in several industrial sectors, such as agrofood, beverages, pharmaceutics, chemicals, pulp & paper, etc. While spectra can easily provide hundreds of measurements across several wavelengths, only a fraction of the collected spectrum conveys relevant information to predict the property of interest. Therefore, the performance of current models is highly related with the ability to select key wavebands, for which the existence of prior knowledge cannot be always secured. Therefore, several feature selection procedures consisting of variants of interval partial least squares (iPLS) have been proposed. These methodologies are however solely focused on determining the most relevant wavebands and do not attempt to further enhance the prediction capabilities within each interval. On the other hand, standard full-spectrum models are often improved by reducing the spectral resolution, but this operation has not been yet synergistically integrated together with waveband selection. As spectral aggregation can effectively improve modelling performance, a multiresolution selection algorithm that simultaneously selects the most relevant wavebands and their optimal resolution is here proposed. By design, this methodology leads to prediction models that are at least as good as the full-spectrum models. The performance comparison made on simulated data and real NIR spectra of gasoline samples also shows that the proposed methodology outperforms iPLS and its variants based on forward and backward selection of intervals in a statistically significant way. Highlights • MR-iPLS is able to simultaneously select wavebands and optimize their local resolution. • Derived models are bound to be at least as good as the full-spectrum models. • Case study illustrates improved performance of several iPLS algorithms. [ABSTRACT FROM AUTHOR]

Abstract The selection of informative variables in partial least squares (PLS) is important in process analytical technology (PAT) applications in the pharmaceutical industry, for example, the calibration of spectrometers. In the past, numerous approaches have been proposed to select the variables in partial least squares. In this work, a new variable selection method for PLS with the weighted variable contribution (PLS-WVC) to the first singular value of the covariance matrix for each PLS component is proposed. Several variants of PLS-WVC with different weighting factors are proposed. One variant of PLS-WVC is equivalent to the PLS with variable importance in projection (PLS-VIP). However, the variants with the correlation between X γ w γ and Y γ q γ as the weighting factor are preferred based on the results of the simulation cases studies. The proposed PLS-WVCs are integrated with interval PLS (iPLS) further to select the informative wavelength intervals for spectroscopic modelling. The utility of the proposed WVC based variable selection methods in PLS is demonstrated with the real spectral data sets. [ABSTRACT FROM AUTHOR]

signals of different shapes and intensities pose a major challenge for the quantitative analysis of mixture spectra from Raman-, Mid-IR- or NMR-spectroscopy. Commonly used approaches for the treatment of such background signals are the subtraction of the background before the spectral analysis, derivative methods and the modeling of the background during the spectral analysis. Only the latter enables the unambiguous distinction between background and spectral features without any distortion of the spectral features. Indirect Hard Modeling (IHM) allows for such joint modeling of spectral features and background signals. However, so far, complex background signals containing distinctive peaks remain challenging for IHM. Furthermore, the user chooses and parametrizes the background treatment method for IHM on a case-by-case basis depending on the spectral properties (e.g. shapes and intensities of the backgrounds or signal-to-noise-ratio). We present a new method called User-independent Nonlinear Modeling using Adjusted Spline-interpolated Knots (UNMASK) that is parametrized by minimizing the prediction error of the calibration spectra, allowing to omit any user-decisions. Besides, unlike other background treatment methods, UNMASK is suited for many cases regardless of the spectral properties, and can even be applied for mixtures with complex background signals, e.g. caused by unknown/unspecified components. We validate UNMASK by applying it to three Raman mixture spectra, covering different application scenarios with varying spectral properties. By comparing the results from UNMASK with different commonly used methods, we show that UNMASK is the only method that consistently leads to low prediction errors for all application scenarios. • New baseline model for Indirect Hard Modeling using spline interpolated knots. • Modeling arbitrarily shaped background signals regardless of properties of spectra. • Parametrization of baseline model during calibration by minimizing prediction error. • User-independent and fully automated. • Low prediction errors for wide range of applications. [ABSTRACT FROM AUTHOR]

A Vis-NIR (visible-near infrared) spectroscopy based method for the end-point determination of the column chromatographic processes of Ginkgo biloba was established. In-line collected Vis-NIR spectra were used in conjunction with various multivariate data analysis methods, including moving block standard deviation (MBSD), principal component scores distance analysis (PC-SDA), and principal component analysis-moving block standard deviation (PCA-MBSD) for the end-point determination of the elution process. Compared with the results validated by high performance liquid chromatography (HPLC) determination, PCA-MBSD was chosen as the most suitable method. The presented method provided an alternative solution to the end-point of the column chromatographic processes, which in the long term depends upon the workers' operating experience traditionally. [ABSTRACT FROM AUTHOR]

In this study, a new strategy called variable importance kernel PLS (VIKPLS) method is developed for near infrared spectral analysis. The wavelength variable importance is incorporated into KPLS by modifying the primary kernel matrix, and variables in the kernel matrix are given the different importance, which provides a feasible way to differentiate between the informative and uninformative variables. The importance of variables is determined by the frequency of variables appearing in the best performing sub-models based on the weighted bootstrap sampling. The performance of VIKPLS is investigated with three real near infrared(NIR) spectroscopic datasets. Examples are given specifically for modifying the linear kernel and Gaussian kernel. Compared with standard kernel PLS, the results show the proposed method can improve the training and prediction performance of KPLS by using variable importance kernel. VIKPLS could be considered as a general and promising mechanism to introduce extra information to improve the performance of KPLS. [ABSTRACT FROM AUTHOR]

In high-dimensional data modeling, variable selection methods have been a popular choice to improve the prediction accuracy by effectively selecting the subset of informative variables, and such methods can enhance the model interpretability with sparse representation. In this study, we propose a novel group variable selection method named ordered homogeneity pursuit lasso (OHPL) that takes the homogeneity structure in high-dimensional data into account. OHPL is particularly useful in high-dimensional datasets with strongly correlated variables. We illustrate the approach using three real-world spectroscopic datasets and compare it with four state-of-the-art variable selection methods. The benchmark results on real-world data show that the proposed method is capable of identifying a small number of influential groups and has better prediction performance than its competitors. The OHPL method and the spectroscopic datasets are implemented and included in an R package OHPL available from https://ohpl.io . [ABSTRACT FROM AUTHOR]

Due to the fluorescence or instrument error, the captured spectra are influenced by the baseline, and the baseline may impact the qualitative and quantitative analysis of spectra. Currently, numerous studies have attempted to remove the impact of spectra baseline, among which multi-scale analysis and regression methods are the most representative methods. It is worth noting that deep learning has made progress in data analysis, but remains rare for baseline correction. In this paper, a baseline correction method based on deep learning is proposed. Specifically, we use the mathematical and physical significance of multi-scale analysis and regression to design a deep learning model and propose a non-convex and non-smooth loss function. Furthermore, some deep learning modules are added to the designed network model. The experiments of simulation, real, and application demonstrate the proposed method achieves state-of-the-art performance. • A baseline correction method based on deep learning, multi-scale analysis and regression. • A CNN-based network architecture is designed via multi-scale analysis. • A non-convex and non-smooth loss function is proposed. • The proposed method has been applied to chemometrics successfully. [ABSTRACT FROM AUTHOR]

In April 2022, the Vistamilk SFI Research Centre organized the second edition of the "International Workshop on Spectroscopy and Chemometrics – Applications in Food and Agriculture". Within this event, a data challenge was organized among participants of the workshop. Such data competition aimed at developing a prediction model to discriminate dairy cows' diet based on milk spectral information collected in the mid-infrared region. In fact, the development of an accurate and reliable discriminant model for dairy cows' diet can provide important authentication tools for dairy processors to guarantee product origin for dairy food manufacturers from grass-fed animals. Different statistical and machine learning modelling approaches have been employed during the workshop, with different pre-processing steps involved and different degree of complexity. The present paper aims to describe the statistical methods adopted by participants to develop such classification model. [ABSTRACT FROM AUTHOR]

In this paper we propose the use of incremental learning for creating and improving multivariate analysis models in the field of chemometrics of spectral data. As main advantages, our proposed incremental subspace-based learning allows creating models faster, progressively improving previously created models and sharing them between laboratories and institutions without requiring transferring or disclosing individual spectra samples. In particular, our approach allows to improve the generalization and adaptability of previously generated models with a few new spectral samples to be applicable to real-world situations. The potential of our approach is demonstrated using vegetable oil type identification based on spectroscopic data as case study. Results show how incremental models maintain the accuracy of batch learning methodologies while reducing their computational cost and handicaps. [ABSTRACT FROM AUTHOR]

The first steps are reported in the development of a new ultrasensitive electrochemical biosensor for detection of Escherichia coli ( E. coli ) in water. Two gold electrodes in a sandwich flow-cell were modified with E. coli polyclonal antibody and exposed to three different concentrations of E. coli . Electrochemical Impedance Spectroscopy was used in combination with Singular Value Decompostion of the complex numbers to monitor the interactions at the electrode surfaces. A linear regression line in the concentration range 10–1000 CFU*ml -1 was obtained without use of redox probes or metal nanoparticles for signal amplification. [ABSTRACT FROM AUTHOR]

A nonparametric and unsupervised method of automatic threshold selection to eliminate noise for spectroscopy data processing is described in this paper. A detecting scheme, named bi-trapezoid criteria, is devised where the threshold is selected according to the turning corner present on the uprising intensity trace. This selecting procedure is very simple, utilizing only basic summation on the sorted intensity sequence to optimize threshold for distinguishing between noises and signals. This approach is effective in selecting appropriate value to filter noise when the distribution of noises is not preconditioned or the gap between signals and noises is not obvious. Testing on both artificial and authentic data under specific quantitative evaluation condition shows that this new method performs better than previous ones. [ABSTRACT FROM AUTHOR]

The combination of spectroscopic measurements and multivariate calibration techniques (chemometrics) has become a state-of-the-art technology for process analytical chemistry. Changes, intended or unintended, in the environmental conditions, the measurement setup or of the measured substance itself can result in a calibration model that is no longer adequate for the intended purpose. In such a situation, either a new model needs to be developed or (calibration) transfer methods, can be applied to transfer models from the original (main, master) to the new (remote, slave) setting. In this contribution, we introduce, discuss and evaluate a wide-ranging subset of transfer approaches available in chemometrics and the field of machine learning, where we focus on techniques applicable in situations where transfer standards, i.e. a set of samples measured under the original as well as the new setting, cannot be provided and only few reference measurements are available for the new setting. The introduced techniques are evaluated on a public data set as well as a new industrial data set displaying three forms of transfer problems. The efficiency of the proposed transfer approaches in terms of the number of required reference measurements compared to full model recalibration can bee confirmed. Average rank maps are presented to provide guidance on a proper choice among evaluated techniques. [ABSTRACT FROM AUTHOR]

The aim of this work is to quantify the relationship between different methods of artificial oil alteration as well as engine oils collected from a passenger car using FTIR (Fourier-transform infrared) spectroscopic data and chemometric methods. We propose a comprehensive procedure for the analysis of FTIR spectra: First, a reconstruction error based pre-processing to filter non-informative variables is introduced, then simultaneous variable selection and parameter estimation using the (weighted) LASSO is performed. Eventually, post-selection inference is applied to derive confidence intervals for the selected model coefficients. The proposed pre-processing methods do not rely on manual selection of FTIR absorption bands suitable for analysis but perform filtering of non-informative variables objectively. With weighted LASSO, experts' knowledge can be integrated with the model. This pipeline for the analysis of FTIR spectroscopic data is demonstrated and validated on a real-world dataset including series of FTIR spectra of used and artificially altered engine oils. • Enhanced model performance by combination of statistical approach and expert knowledge. • Identification of non-informative variables by automatic filtering procedure. • Quantitative relation between different degradation pathways. [ABSTRACT FROM AUTHOR]

In this paper, a two-level independent component regression (ICR) model is developed for multivariate spectroscopic calibration. Compared to the traditionally used principal component regression and partial least squared regression model, the ICR model is more efficient to extract high order statistical information from the spectra data. To improve the calibration performance, an ensemble form of the ICR model is proposed. In the first level of the method, various subspaces are constructed based on the independent component decomposition of the original data space. Meanwhile, by defining a related index, the most important variables in each subspace are selected for ICR modeling, which form the second level of the proposed method. A Bayesian inference strategy is further developed for probabilistic combination of calibration results obtained from different subspaces. For performance evaluation, two case studies are carried out on a benchmark spectra dataset. [ABSTRACT FROM AUTHOR]

This study aims at resolving the spectral profiles of pure components from the mixture THz data. A method based on simple-to-use interactive self-modeling mixture analysis (SIMPLISMA), one of self-modeling curve resolution (SMCR) techniques, was put forward to deal with two-way THz spectral data of unknown mixtures for spectral resolution of single species. SMCR requires pure variables which have intensity contributions from only one component for extracting the content information. However, for the spectra of many materials in THz region (about 0.1–3 THz), pure variables are usually unavailable for the overlap of absorption peaks or the presence of baseline. To settle this problem, inverted second derivative was employed as an intermediate step to optimize the determination of pure variable. To test the effect, the proposed method was performed on simulated and experimental THz spectra of multicomponent systems. It is demonstrated that SIMPLISMA combined with inverted second derivative spectra is a promising tool to deconvolve multicomponent systems contributing to the identification of ingredients in unknown mixture in THz region. [ABSTRACT FROM AUTHOR]

While partial least squares (PLS) and principal component regression (PCR), the most popular regression techniques in chemometrics, may theoretically be able to deal with large numbers of possibly correlated variables, as occurring in the analysis of spectroscopic data, the importance of performing some form of variable selection in practical applications has been widely discussed and acknowledged. In this work we address this problem via proposing a sparse regression algorithm, referred to as fused stagewise regression (FSR), which iteratively performs a selection of connected regions of variables (wavelengths), while being quite easy to implement and interpret, due to its resemblance to typical steps in iterative manual feature selection procedures. We evaluate the proposed variable selection technique on a publicly available benchmark data set and compare the performance of PLS models built on the determined selection to ones obtained by state-of-the-art feature selection methods from the fields of chemometrics and machine learning. In order to ensure robust feature selection, we integrate the individual selection methods into an extensive repeated cross validation procedure. For the data set under investigation, it is shown that FSR performs at least as good as state-of-the-art approaches and well within the range of variable selections provided by experts. [ABSTRACT FROM AUTHOR]

In this work, by analysing excitation–emission matrix fluorescence (EEMF) spectroscopic data sets with multivariate curve-resolution alternating least square (MCR-ALS) analysis under the commonly used mathematical constraint of non-negativity, an attempt was made to understand how the outcome of curve resolution technique can be influenced by the design of calibration sets. MCR-ALS analysis with commonly used mathematical constraints has been successfully used to analyse multi-component systems having significant spectral overlap. However, the present work showed that even for a simple system where the components do not have spectral overlap with each other, the MCR-ALS technique can resolve the spectra only if the calibration set is suitably designed. To carry out this study, three fluorophores—benzo[ a ]pyrene, perylene, and pyrene—having no spectral overlap with each other, were chosen. Selection of such fluorophores would enable the observation of fluorescence signatures of one fluorophore if it appears in the MCR-ALS retrieved profile of other fluorophores. Ten calibration sets with different approaches were created. EEMF data sets acquired for these calibration sets were subjected to MCR-ALS analysis. With most of the calibration sets, it was difficult to retrieve the pure EEMF spectra of all three fluorophores even if they did not have spectral overlap with each other. This work also evolved certain criterion for creating the calibration set which would enable the retrieval of pure EEMF spectra of all fluorophores of a multifluorophoric mixture using curve resolution technique. The present work clearly shows that pure spectral profiles of all the fluorophores under commonly used mathematical constraints can only be retrieved provided curve resolution analysis is performed on a suitably designed calibration set. [ABSTRACT FROM AUTHOR]

A new method for abnormal spectrum detection based on the mixed model of samples is proposed. The method can detect abnormal spectra on the condition that the content values are unknown. The method consists of four steps. Firstly, mixed vector of the prediction sample is calculated according to the mixed model of samples. Secondly, estimated spectrum of the prediction sample is calculated according to the mixed ratio and the spectrum of calibration samples. Thirdly, the difference between the estimated spectrum and the measuring spectrum is calculated. Lastly F-statistical test is carried out to detect the abnormal spectrum according to the variance. The method is compared with the MMS and PLS algorithms. In the experiment, it is assumed that the contents of the prediction samples are unknown for the new method. For MMS and PLS, the contents of the prediction samples are known, and when the prediction error is bigger than three times the root mean square error of prediction (RMSEP), the spectrum is identified as abnormal spectrum. Results from calculations show that the new method has better detection performances for abnormal spectrum caused by measurement background changes, instrumental noise increase, and the condition of detection samples containing non-calibration content than MMS and PLS algorithms. The new method provides a new approach to detect the spectrometer performance including the background changes and noise increase in advance. [ABSTRACT FROM AUTHOR]

In order to successfully apply differentiation to quantitative analytical applications, it is important to develop a method of differentiation with noise immunity to optimize the signal-to-noise ratio. Therefore, a direct method for the approximate fractional order differentiation (FOD) of a noised signal is proposed. Using the proposed method, arbitrary order differentiation of a signal can be realized by the convolution between the fractional order differentiation of the Gaussian function and the noised signal, and the integer order differentiation is just a special case of the FOD. The proposed method is verified by comparing the FOD of the noise and noise-free peak-signals. Moreover, the proposed method is used to enhance a mass spectrum signal and locate the position of each peak. [ABSTRACT FROM AUTHOR]

Similarity-measure-based spectrum matching is an effective approach to chemical compound identification. When the sizes of both the query library and the reference library become increasingly large, most existing spectrum-matching methods encounter a seriously heavy computation burden. In this paper, an effective and efficient compound-identification approach is proposed based on the frequency features of mass spectra. Considering the sparsity of mass spectra, a nonzero feature-selection strategy is proposed to decrease the feature dimensionality of mass spectra. To further improve its efficiency, a correlation-based filtering strategy is presented to select the most correlated reference spectra in order to create a reduced reference library. Based on the decreased features and the reduced reference library, the frequency-feature-based composite similarity measures are computed to estimate the chemical abstracts service (CAS) registry numbers of the mass spectra blue in a query library. Due to the reduction in both the feature dimensionality and the reference library, the computation time of the proposed method is only about 6%–11% of that of the existing methods, while the identification performance remains sufficiently competitive. Experimental results demonstrate the feasibility and efficiency of the proposed method. [ABSTRACT FROM AUTHOR]

In the analysis of spectrum signals, it is especially important to identify the peaks of a spectrum signal. Due to various noise, spectrum signals must be smoothed in advance, while preserving peaks information. In this paper, we propose a peak-aware guided filtering (PAGF) for peak-preserving smoothing of a noisy spectrum signal. The content of a guidance signal is taken into account in the filtering process, which can provide smooth structure information and peak position of the input signal. The guidance signal is generated from the input signal by a nonlinear diffusion filtering (NDF) since the output of NDF provides sufficiently smooth structure and accurate peak position for the input signal. The proposed PAGF is validated by experiments on commonly tested signals, in comparison to previous peak-preserving smoothing methods. The results show that the PAGF can better preserve the peak height, width and position of the input spectrum signal while removing the noise. • Propose peak-aware guided filter for peak-preserving smoothing of spectrum signal. • Our methodology integrates popular guided image filtering and diffusion filtering. • Our method extends the application of guided filteringto spectrum signal denoising. [ABSTRACT FROM AUTHOR]

In the analysis of spectrum signals, it is especially important to identify the peaks of a spectrum signal. Due to various noise, spectrum signals must be smoothed in advance, while preserving peaks information. In this paper, we propose a peak-aware guided filtering (PAGF) for peak-preserving smoothing of a noisy spectrum signal. The content of a guidance signal is taken into account in the filtering process, which can provide smooth structure information and peak position of the input signal. The guidance signal is generated from the input signal by a nonlinear diffusion filtering (NDF) since the output of NDF provides sufficiently smooth structure and accurate peak position for the input signal. The proposed PAGF is validated by experiments on commonly tested signals, in comparison to previous peak-preserving smoothing methods. The results show that the PAGF can better preserve the peak height, width and position of the input spectrum signal while removing the noise. • Propose peak-aware guided filter for peak-preserving smoothing of spectrum signal. • Our methodology integrates popular guided image filtering and diffusion filtering. • Our method extends the application of guided filteringto spectrum signal denoising. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, a support vector regression algorithm in the sum of reproducing kernel Hilbert spaces (SVRSS) is proposed for multivariate calibration. In SVRSS, the target regression function is represented as the sum of several single kernel decision functions, where each single kernel function with specific scale can approximate certain component of the target function. For sum spaces with two Gaussian kernels, the proposed method is compared, in terms of RMSEP, to traditional chemometric PLS calibration methods and recent promising SVR, GPR and ELM methods on a simulated data set and four real spectroscopic data sets. Experimental results demonstrate that SVR methods outperform PLS methods for spectroscopy regression problems. Moreover, SVRSS method with multi-scale kernels improves the single kernel SVR method and shows superiority over GPR and ELM methods. [Copyright &y& Elsevier]

Abstract: Field mobile mass spectrometer is pivotal apparatus for real-time qualitative and quantitative analyses of chemical substances in situ environment pollution detection. To solve spectrum peak signal interfered by complicated noise, and to recognize irregular peak shape as well as quick monitoring, a real-time denoising and hyper-accuracy peak identification integrative approach for field mobile mass spectrometer using lifting-based wavelet transform (LWT) and Gaussian model has been developed. First, LWT was applied to eliminate the noise and to search for mass peak parameters in raw spectral peak data. Then, fitting the irregular mass peaks with Gaussian multi-peaks, a regular spectrum signal was obtained for further processing. Both of synthetic and apparatus experiment results show that LWT is a fast and effective denoising and peak identification method and retained the original peak features. The denoising effect (SNR/RMSE) by LWT was superior to Savitzky–Golay method used widely by experimental mass spectrometer, and the processing time was shortened obviously. Moreover integrated with Gaussian fitting algorithm, the peak parameters (the peak area A, centroid c, and half peak's width w) had been optimized. As the result, qualitative and quantitative accuracies of FMMS increased consequently. In addition, the approach achieved data compression. [Copyright &y& Elsevier]

Abstract: In this study, taking advantage of multi-resolution property of wavelet transform along with a simple statistics, a careful search method for feature selection is proposed. This method enables identifying a set of wavelet coefficients as features that maximizes the discrimination of chemical spectra in different classes. In spite of variety of application of discrete wavelet transform (DWT) and wavelet packet transform (WPT) in manifesting the information content of signal, continuous wavelet transform (CWT) is used in this approach. Continuous wavelet transform due to its redundancy property is able to provide a finer multi-resolution space that visualizes all information content of spectra. The proposed method is tested on three spectral data sets consisting of 1H NMR analysis of table wines, near infrared absorbance spectra of meat samples and NIR diffuse transmission spectra of pharmaceutical tablets. Using the suggested method a small number of features are selected and fed as an input to the different classifiers, Partial Least Squares Discriminant Analysis (PLS-DA), K-Nearest Neighbor (K-NN) and Soft Independent Modeling of Class Analogy (SIMCA). It is confirmed that the classification models developed on the selected wavelet coefficients yield higher correct classification rate (CCR) than the models constructed with original features which confirm the adequacy of the generated feature space for classification. [Copyright &y& Elsevier]

Abstract: Terahertz time domain spectroscopy (THz-TDS) has been utilized as an available tool for the identification and analysis of compounds owing to its unique characters. The THz spectrum of nitrofen was detected and the comparison with other two solid pesticides (thiacloprid and acetamipyrid) was provided to demonstrate pesticide's unique characteristics in THz region. For the quantitative analysis of pesticide, we applied THz-TDS to detect nitrofen mixtures with different weight ratios, and employed partial least squares (PLS), interval PLS (iPLS), moving window PLS (mwPLS) and backward interval PLS (biPLS) algorithms to achieve the best analysis accuracy through a variable selection. The regression models were optimized by root mean square error of cross validation (RMSECV) in the calibration set, and the performance of models was evaluated according to the root mean square error of prediction (RMSEP) and correlation coefficient (R). Compared with the results of the other three methods, the biPLS model acquired the best performance with RMSEP=0.4064, R =0.9995. [Copyright &y& Elsevier]

Abstract: An efficient method to extract and store information from NMR spectra is proposed that is suitable for comparison and construction of a search engine. This method based on trees doesn't require any peak picking or any pre-treatment of the data and is found to outperform the currently available methods, both in terms of compactness and velocity. Our approach was tested for 1D proton spectra and 2D HSQC spectra and compared with the method proposed by Pretsch and coworkers [1,2] [Bodis et al. 2007, Bodis et al. 2009]. Additionally, the correspondence between spectral and structural similarity was evaluated for both methods. [Copyright &y& Elsevier]

Abstract: Regions of explanatory variables, X, are attempted to be selected in many fields such as spectral analysis and process control. A genetic algorithm-based wavelength selection (GAWLS) method is one of the methods used to select combinations of important variables from X-variables using regions as a unit of measurement. However, a partial least squares method is used as a regression method, and hence, a GAWLS method cannot handle nonlinear relationship between X and an objective variable, y. We therefore proposed a region selection method based on GAWLS and support vector regression (SVR), one of the nonlinear regression methods. The proposed method is named GAWLS–SVR. We applied GAWLS–SVR to simulation data and industrial polymer process data, and confirmed that predictive, easy-to-interpret, and appropriate models were constructed using the proposed method. [Copyright &y& Elsevier]

Abstract: Within the framework of multi-block data sets, multi-block principal component analysis has been successfully used as a tool to investigate the structure of spectroscopy, -omics and sensory data. The determination of the successive principal components involves a deflation procedure which can be performed according to several strategies. We discuss the respective interest of these strategies and show orthogonality properties related to the vectors of loadings or to the scores. Reconstruction formulas for the data blocks are established for each deflation strategy. Interpretational aspects of the different deflation strategies are discussed and illustrated on the basis of a real and a simulated data set. [Copyright &y& Elsevier]

Abstract: A spectroscopic calibration model is only valid to predict samples within the current calibration sample space span. This space is characterized by the calibration spectral variances set by the sample matrix properties and instrument environment (the primary conditions). Prediction samples commonly have new spectral variances (the secondary conditions) and dynamic model maintenance is needed. Previous work has shown that variants of Tikhonov regularization (TR) are capable of accomplishing this task by updating the primary model with only a few secondary condition samples (a current standardization set). An aspect of the TR variants is a weight (tuning parameter) for the small standardization set augmented to the full primary calibration sample set. In past work, this tuning parameter in combination with a second regularization parameter is graphically assessed to select a single updated model. Developed in this paper is a novel graphical consensus approach that selects a family of models across a range of tuning parameter values. Thresholds on model merit values are used to identify appropriate updated models. Model merits can be R2, intercept, and slope for the primary calibration and standardization sets, root mean square error (RMSE) terms, and/or the model vector magnitude. Two previously used TR variants are studied. One TR modification requires analyte reference values and the other variation uses no reference values. The TR consensus approach with reference samples is applied to updating a laboratory based near infrared (NIR) primary calibration model to predict active pharmaceutical ingredient (API) tablet concentrations from NIR spectra measured on tablets produced in the full production secondary conditions. The TR approach without reference samples updates a NIR pure component analyte model at one temperature to new temperature dependent sample matrix conditions. In both studies, consensus models predict equivalently to individual models selected by previously developed graphical approaches. The consensus approach is also applied to model updating by partial least squares (PLS). While PLS and TR predict similarly, PLS can be limited by the discrete factorization process. The described consensus approach is also applicable to just primary calibration. [Copyright &y& Elsevier]

Abstract: Out-of-scope sample detection is a key element in successful pharmaceutical quality monitoring by multivariate models and spectroscopic measurements. It is of importance that the spectral data for new measurements are spectrally equivalent to that used in developing the calibration set to ensure that the predicted values will be accurately expressed. In the present work we evaluated prediction uncertainty approaches as applied to partial least squares (PLS) models that were applied to a pharmaceutical tablet assay method incorporating near infrared (NIR) spectroscopy. During the method implementation with TQ Analyst software, the algorithm was compared with the Unscrambler software, the SIMCA-P+ software, Error-in-Variable (EIV) approach, and verified by the use of PLS_Toolbox software. It is found that the uncertainty values are very close between TQ Analyst and the Unscrambler though the algorithms are quite different. The magnitude of the uncertainty can be sensitive to variations in instrument performance and tablet composition. Confidence limit setting for the prediction uncertainty is discussed with consideration of confidence limits used in Hotelling-T2 and Q-residual statistics and multiple linear regression. [Copyright &y& Elsevier]

Abstract: The key point in multivariate calibration is to build an accurate regression relationship between the predictors and responses. In this paper, we first use extreme learning machine (ELM) to build spectroscopy regression model. Then, we propose a combinational ELM (CELM) method in which the decision function is represented as a sum of a linear hidden-node output function (activation function) and a nonlinear hidden-node output function. As the output functions map the input spectral signal to linear and nonlinear feature spaces respectively, the proposed method can effectively describe the linear and nonlinear relations existed in spectroscopy regression by the CELM output weights vector which can be simply resolved by ridge least squares or alternative iterative regularization. The proposed method is compared, in terms of RMSEP, to PLS and ELM on simulated and real NIR data sets. Experimental results demonstrate the efficacy and effectiveness of the proposed method. [Copyright &y& Elsevier]

Abstract: In this paper a calibration transfer method is used to achieve bilinearity for augmented first order kinetic data. First, the proposed method is investigated using simulated data and next the concept is applied to experimental data. The experimental data consists of spectroscopic monitoring of the first order degradation reaction of carbaryl. This component is used for control of pests in fruits, vegetables, forages, cotton and other crops. It is highly toxic and likely human carcinogen, and is lethal to many non-target beneficial insects. The kinetic experiment is performed at different pH-values and emission wavelengths using an excitation wavelength equal to 275nm. Rate constants of different data matrices at different pH values were calculated based on a hard modeling method. Analysis of simulated and experimental data shows that if there is a deviation from bilinearity, applying the model based methods to augmented datasets leads to inaccurate results. The application of a calibration transfer method as an additional step in the hard modeling procedure improves the results, and accurate estimation of reaction rate constants are obtained. The proposed method was compared to Local Spectra Mode of Analysis (LSMA) which was proposed by Puxty et al. A comparison of the results shows that the proposed method is more efficient than LSMA and leads to less uncertainty in estimated rate constants and less percent error in the relative residuals. [Copyright &y& Elsevier]

Abstract: In order to valorize olive stones we are studying their dilute-acid hydrolysate (DAH) composition, and trying to highlight the effect of the overliming process on DAH composition in order to perform an effective treatment that maximize sugars concentrations (xylose) and minimizes the amount of toxic materials (FF and HMF). An 2231 experimental design was established to describe the effects of three controlled factors with distinct levels: pH (10 and 12), temperature (25 and 60°C) and detoxification time (15, 30 and 60min) on the concentration of xylose, FF and HMF. A better understanding of this overliming process was possible by exploring the chromatograms obtained with 2D Correlation Spectroscopy (2DCOS). 2D correlation spectroscopy gave information about the relations that exist between chromatographic peaks and chromatograms. The order in which the constituents vary can be deduced from the sign of peaks in the synchronous and asynchronous matrices, facilitating the interpretation of kinetic studies. [Copyright &y& Elsevier]

Abstract: Hyperspectral imaging in various spectral domains is used to perform in situ biochemical analysis of biological material. In the present work, two complementary microspectroscopies—mid-infrared and fluorescence—were coupled to characterise maize stem cell walls. In practice, each microspectroscopy operates with its own resolution, infrared pixels covering a 5×5μm² surface and fluorescence pixels a 1×1μm² area, making the coupling of both spectral ranges through unfolding the spectral images not straightforward. A data structure that preserved the spatial resolution in the two domains was built resulting in a two-way data table with a spatial way and a spectral way, paired to a three-way data table with two spatial ways and a spectral way. The multivariate inter-battery Tucker analysis was investigated in order to give each spectral domain a symmetric role. First, the inter-battery Tucker analysis was applied after averaging the three-way data table under the third mode to obtain a classical two-way data table. This led to work at the lowest resolution. Second, the method has been extended for coupling the three-way data table with the two-way data table. The spectral loadings were similar in both cases showing that phloem cell walls were rich in hydroxycinnamic acids and sclerenchyma cell walls in lignins. The extended version provided with fluorescence scores that preserved the spatial resolution showing more fluorescence intensity in cell junctions. The strong point of the extended inter-battery Tucker analysis is to allow the joint analysis of the two tables without altering their qualities. [Copyright &y& Elsevier]

Abstract: Although there are some distinctive peaks in mid-IR region of the electromagnetic spectrum for carbonyl compounds, it is very difficult to assign a FT-IR spectrum to a specific carbonyl functional group due to the presence of other functional groups, which change the position of the distinctive peaks. Here, we analyzed the FT-IR spectra of a large set of carbonyl compounds by chemometrics methods to differentiate between the different carbonyl functional groups. FT-IR spectra of 370 carbonyl compounds (149 carboxylic acids, 47 aldehydes, 110 esters and 64 ketones) were collected from the Spectral Database for Organic Compounds and then were converted to digital data using a home-made program. The extended canonical variate analysis combined with partial least squares discriminate analysis method (ECVA-PLS-DA) was employed as a supervised classification method. Classification analysis by ECVA-PLS-DA resulted in a suitable classification model such that one can discriminate between the different carbonyl compounds using FT-IR spectra with a small error. The classification errors (reported as percentage of misclassified compounds) were 1.8% and 7.8% for training and test sets, respectively. Considering high structural diversity of the studied compounds and the employment of different methods for acquiring the spectra (i.e., KBr disk, CCl4 solution, liquid film and Nujol moll) there are acceptable errors. Thus, it is concluded that with the help of chemometrics methods, one is able to differentiate the carbonyl compounds using their IR spectra without need to extra spectroscopic information. This can be considered as a significant improvement in structural characterization of organic compounds using only IR spectroscopy. [Copyright &y& Elsevier]

Abstract: Soft modeling (SM) methods can be used to resolve spectroscopic data from complicated reaction processes with unknown kinetics, with the exception of data containing a component for which there is no spectroscopic information available, such as an intermediate that does not absorb in the UV–visible region. In this work, modified SM methods were developed to resolve these undetectable components. Based on the mass balance principle, the mass balance error (MBE) method was first applied to determine whether the undetectable component existed. Next, the evolving error analysis (EEA) method was developed to search the local mass balance region (LMBR) where the concentration of the non-absorptive component was low enough to be neglected. In the LMBR, the concentration profiles of all absorptive components were scaled according to least squares regression. Subsequently, more reliable results were obtained using the evolving time region iteration (ETRI) method. Based on the mass balance principle, the concentration profile of the undetectable component was resolved for the entire time period. Both simulated and experimental data from an autocatalytic reaction were used to demonstrate the feasibility of the proposed method. In the autocatalytic oxidation of sodium oxalate by acidic potassium permanganate, the product Mn(II) was determined to be non-absorptive. Using the methods described above, the pure spectra of three other absorptive components and the scaled concentration profiles of four Mn species, including two intermediates, were all resolved. As a result, the mechanism of the reaction was more clearly described. [Copyright &y& Elsevier]

Abstract: Baseline correction is an important pre-processing technique used to separate true spectroscopic signals from interference effects or remove background effects, stains or traces of compounds, e.g. in 2D gel electrophoresis. In some cases parts of the spectra or images need correction using rigid baselines (limited curvature) while other parts need more flexible baselines (more curvature). Often one has to make a compromise that is not optimal over the whole spectral range, or focus on one part and let the rest be treated sub-optimally. A customizing wrapper is proposed that rescales the spectrum abscissa and therefore makes the baseline correction algorithm behave right in all parts of the spectra. Improvements are demonstrated both visually and through regression using recently reported Raman spectra on melted fat from pork adipose tissue. [Copyright &y& Elsevier]

Abstract: Nowadays, image analysis is becoming more important because of its ability to perform fast and non-invasive low-cost analysis on products and processes. Image analysis is a wide denomination that encloses classical studies on gray scale or RGB images, analysis of images collected using few spectral channels (sometimes called multispectral images) or, most recently, data treatments to deal with hyperspectral images, where the spectral direction is exploited in its full extension. Pioneering data treatments in image analysis were applied to simple images mainly for defect detection, segmentation and classification by the Computer Science community. From the late 80s, the chemometric community joined this field introducing powerful tools for image analysis, which were already in use for the study of classical spectroscopic data sets and were appropriately modified to fit the particular characteristics of image structures. These chemometric approaches adapt to images of all kinds, from the simplest to the hyperspectral images, and have provided new insights on the spatial and spectroscopic information of this kind of data sets. New fields open by the introduction of chemometrics on image analysis are exploratory image analysis, multivariate statistical process control (monitoring), multivariate image regression or image resolution. This paper reviews the different techniques developed in image analysis and shows the evolution in the information provided by the different methodologies, which has been heavily pushed by the increasing complexity of the image measurements in the spatial and, particularly, in the spectral direction. [Copyright &y& Elsevier]

Abstract: Based on near infrared spectroscopy, a method for rapid identification of crude oil type, named moving window correlation coefficient method, was proposed. This new spectral searching method can distinguish highly similar crude oil spectra. The main parameters of the moving window correlation coefficient method were discussed and selected in this study. Combined with crude oil assay database, detailed crude oil assay of unknown crude oils can be rapidly provided by this method. With the expansion of the NIR spectral library and assay database, this strategy would play more and more important role in making business decisions, and in planning, controlling and optimizing operations of refineries. [Copyright &y& Elsevier]

Abstract: In this paper, by using the Haar wavelet as a tool, we design a fractional-order differential filter to develop the fractional-order derivative spectroscopy (FDS), and then we use it to resolve the overlapped Lorentzian peaks. Resolutions are directly characterized by spectral parameters extracted from the overlapped peaks. Comparison results between our method and the Peakfit software indicate the FDS can be used to resolve overlapped Lorentzian peaks effectively and satisfactorily. [Copyright &y& Elsevier]

Abstract: This work investigates the potential of using simultaneously near infrared (NIR) and mid infrared (MIR) spectroscopy for the quantification of vegetable oil in diesel/biodiesel blends. The two spectral ranges were used separately with PLS regressions. To combine the two pieces of information, first a concatenated matrix was used and then H-PLS and S-PLS models were constructed. The models were compared in term of prediction errors (RMSEP). The results obtained in NIR spectroscopy were better than the ones obtained with MIR spectroscopy (considering the influence of the pretreatment and of the selected variables range). The multiblock methodology seems to be of great interest in quantitative analysis with the simultaneous use of information from the MIR and NIR spectroscopy. [Copyright &y& Elsevier]

Abstract: Demand for a renewable liquid-phase fuel has reinvigorated interest in the development of novel biofuels from algae. The effects of fluctuating nutrients, temperature, and predation which arise in algal aquaculture necessitate novel analytical methods to assess and optimize the health of these cultures. In this article we describe the use of trilinear modeling to these ends, using imaging data of endogenous fluorescence from the marine diatom, Thalassiosira pseudonana, collected on a hyperspectral fluorescence confocal microscope (HSF-CM). By combining the spatial and spectral data with the additional physical parameter introduced by the varying photobleach rates of endogenous fluorophores, we obtain a means for employing trilinear analysis in this system. In this case, analysis was implemented using the method of parallel factors analysis (PARAFAC) which alleviates the rotational ambiguities involved in multivariate analysis of 2-way data. By imposing a nutrient limitation in our diatom culture, and collecting single cell microscopy data at various time points during the nutritional stress, PARAFAC results reveal evolution of the spectral emission characteristics of six components and a background as a result of incremental photobleaching. Although the underlying mechanisms for these observations are not addressed, the differences observed between nutritively normal and nutritively stressed cells provide signatures for early detection of health deterioration in the algal culture. Finally, the combination of hyperspectral fluorescence microscopy with PARAFAC is generally applicable for the study of complex biological systems with endogenous fluorescence. [Copyright &y& Elsevier]

Abstract: This work tackles the problem of convolution of femtosecond spectrokinetic data with the instantaneous response function (IRF) of the measurement by hybrid hard- and soft-modeling multivariate curve resolution (MCR). Fitting the time-dependent concentration profiles obtained during the MCR iterative optimization by a kinetic model convoluted with a parametric description of the IRF, the approach benefits of the advantages of both hard- and soft-modeling. The performance of the method is first proven on synthetic data sets with overlap of pure time-dependent concentration and/or spectral profiles and presence of interfering contributions. The method is afterwards applied to the resolution of femtosecond transient absorption spectroscopy data probing ultrafast photoinduced processes implied in the photochromism of anils. An excited-state proton transfer appearing in ~80fs is characterized using an instrumental system with a severe influence of the convoluted instrumental response function, defined by a time resolution of ~150fs. [Copyright &y& Elsevier]

Abstract: Multivariate Curve Resolution (MCR) aims to blindly recover the concentration profile and the source spectra without any prior supervised calibration step. It is well known that imposing additional constraints like positiveness, closure and others may improve the quality of the solution. When a physico-chemical model of the process is known, this can be also introduced constraining even more the solution. In this paper, we apply MCR to Ion Mobility Spectra. Since instrumental models suggest that peaks are of Gaussian shape with a width depending on the instrument resolution, we introduce that each source is characterized by a linear superposition of Gaussian peaks of fixed spread. We also prove that this model is able to fit wider peaks departing from pure Gaussian shape. Instead of introducing a non-linear Gaussian peak fitting, we use a very dense model and rely on a least square solver with L1-norm regularization to obtain a sparse solution. This is accomplished via Least Absolute Shrinkage and Selection Operator (LASSO). Results provide nicely resolved concentration profiles and spectra improving the results of the basic MCR solution. [ABSTRACT FROM AUTHOR]

Abstract: Ultrafiltration (UF) of milk in the dairy industry generates significant quantities of UF permeate as a by-product. In the past decade UF permeate has been utilized as a mixing ingredient in the standardization of milk. Due to the role of permeate as an ingredient, it is important that the variation in composition is known and controlled in order to ensure uniform quality of the standardized milk. In this investigation we evaluate if the composition of permeate from the ultrafiltration of milk varies between different ultrafiltration plants as assessed by infrared spectroscopy. A total of 40 permeate samples from six production sites at different geographical locations were analyzed by infrared spectroscopy. Principal component analysis of the infrared spectra showed that it was possible to classify the plant manufacturer of permeates with great accuracy. Loading spectra revealed that total solids (mainly lactose) and protein were the main sources of compositional variation between the different sites. Through an orthogonalization procedure of the infrared spectra the spectral variation due to total solid and protein content was removed. Neither the unsupervised principal component analysis nor the supervised extended canonical variate analysis could classify the permeate plant origin after orthogonalization of the infrared spectra. The result shows that, besides total solids and protein variation, permeate from the six ultrafiltration sources does not contain plant specific composition in their infrared signature. The study demonstrates how multiple external factors can be removed from the spectral data by orthogonalization. [ABSTRACT FROM AUTHOR]

Abstract: Extraction of peak areas and mass spectral information from chromatography mass spectral data such as obtained in metabolomics measurements requires much effort and the quality is often subjective to the operator that handles the data at hand. In multiple file deconvolution, all samples are processed simultaneously and alignment issues are part of the modeling strategy. However, processing the total data set as a whole is an impossible task and therefore the data processing task requires segmentation. Two intertwined divide and conquer strategies are proposed. The first strategy divides the retention time axis into equal parts and the second strategy divides the total data set into a model and a prediction data set. Dividing the data into smaller segments allows us to conquer the total problem. Post processing of the resulting matrices with peak areas and mass spectra ensures that a matrix with peak areas ready for statistics and a matrix with mass spectral information ready for peak annotation is obtained. The proposed methodology is implemented within a package called TNO-DECO but can easily be implemented in other data pre-processing approaches. [ABSTRACT FROM AUTHOR]