Ludovic Duponchel

Resolution and segmentation of hyperspectral biomedical images by multivariate curve resolution-alternating least squares.

MCR-ALS is a resolution method that has been applied in many different fields, such as process analysis, environmental data and, recently, hyperspectral image analysis. In this context, the algorithm provides the distribution maps and the pure spectra of the image constituents from the sole information in the raw image measurement. Based on the distribution maps and spectra obtained, additional information can be easily derived, such as identification of constituents when libraries are available or quantitation within the image, expressed as constituent signal contribution. This work summarizes first the protocol followed for the resolution on two examples of kidney calculi, taken as representations of images with major and minor compounds, respectively. Image segmentation allows separating regions of images according to their pixel similarity and is also relevant in the biomedical field to differentiate healthy from non-healthy regions in tissues or to identify sample regions with distinct properties. Information on pixel similarity is enclosed not only in pixel spectra, but also in other smaller pixel representations, such as PCA scores. In this paper, we propose the use of MCR scores (concentration profiles) for segmentation purposes. K-means results obtained from different pixel representations of the data set are compared. The main advantages of the use of MCR scores are the interpretability of the class centroids and the compound-wise selection and preprocessing of the input information in the segmentation scheme.

Comparison of supervised pattern recognition methods with McNemar’s statistical test: Application to qualitative analysis of sugar beet by near-infrared spectroscopy

Abstract The application of supervised pattern recognition methodology is becoming important within chemistry. The aim of the study is to compare classification method accuracies by the use of a McNemar’s statistical test. Three qualitative parameters of sugar beet are studied: disease resistance (DR), geographical origins and crop periods. Samples are analyzed by near-infrared spectroscopy (NIRS) and by wet chemical analysis (WCA). Firstly, the performances of eight well-known classification methods on NIRS data are compared: Linear Discriminant Analysis (LDA), K -Nearest Neighbors (KNN) method, Soft Independent Modeling of Class Analogy (SIMCA), Discriminant Partial Least Squares (DPLS), Procrustes Discriminant Analysis (PDA), Classification And Regression Tree (CART), Probabilistic Neural Network (PNN) and Learning Vector Quantization (LVQ) neural network are computed. Among the three data sets, SIMCA, DPLS and PDA have the highest classification accuracies. LDA and KNN are not significantly different. The non-linear neural methods give the less accurate results. The three most accurate methods are linear, non-parametric and based on modeling methods. Secondly, we want to emphasize the power of near-infrared reflectance data for sample discrimination. McNemar’s tests compare classification developed with WCA or with NIRS data. For two of the three data sets, the classification results are significantly improved by the use of NIRS data.

Multivariate curve resolution methods in imaging spectroscopy: influence of extraction methods and instrumental perturbations.

Imaging spectroscopy is becoming a key field of analytical chemistry. In the face of more and more complex samples, we actually need accurate microscopic insight. Nowadays, the methods used to produce concentration maps of the pure compounds from spectral data sets are based on the classical univariate approach although multivariate approaches are sometimes investigated. But in any case, the analytical quality of the chemical images thus provided cannot be discussed since no reference methods are at our disposal. Thus the proposed research focuses on the application of multivariate methods such as Orthogonal Projection Approach (OPA), SIMPLE-to-use Self-modeling Mixture Analysis (SIMPLISMA), Multivariate Curve Resolution - Alterning Least Squares (MCR-ALS), and Positive Matrix Factorization (PMF) for imaging spectroscopy. A systematic and quantitative characterization of the accuracy of spectra and images extraction is investigated on mid-infrared spectral data sets. Of special interest is the influence of instrumental perturbations such as noise and spectral shift on the extraction ability to access the algorithms robustness.

The Organization Pattern of Root Border-Like Cells of Arabidopsis Is Dependent on Cell Wall Homogalacturonan

Border-like cells are released by Arabidopsis (Arabidopsis thaliana) root tips as organized layers of several cells that remain attached to each other rather than completely detached from each other, as is usually observed in border cells of many species. Unlike border cells, cell attachment between border-like cells is maintained after their release into the external environment. To investigate the role of cell wall polysaccharides in the attachment and organization of border-like cells, we have examined their release in several well-characterized mutants defective in the biosynthesis of xyloglucan, cellulose, or pectin. Our data show that among all mutants examined, only quasimodo mutants (qua1-1 and qua2-1), which have been characterized as producing less homogalacturonan, had an altered border-like cell phenotype as compared with the wild type. Border-like cells in both lines were released as isolated cells separated from each other, with the phenotype being much more pronounced in qua1-1 than in qua2-1. Further analysis of border-like cells in the qua1-1 mutant using immunocytochemistry and a set of anti-cell wall polysaccharide antibodies showed that the loss of the wild-type phenotype was accompanied by (1) a reduction in homogalacturonan-JIM5 epitope in the cell wall of border-like cells, confirmed by Fourier transform infrared microspectrometry, and (2) the secretion of an abundant mucilage that is enriched in xylogalacturonan and arabinogalactan-protein epitopes, in which the cells are trapped in the vicinity of the root tip.

Simultaneous data pre-processing and SVM classification model selection based on a parallel genetic algorithm applied to spectroscopic data of olive oils

Classification is an important task in chemometrics. For several years now, support vector machines (SVMs) have proven to be powerful for infrared spectral data classification. However such methods require optimisation of parameters in order to control the risk of overfitting and the complexity of the boundary. Furthermore, it is established that the prediction ability of classification models can be improved using pre-processing in order to remove unwanted variance in the spectra. In this paper we propose a new methodology based on genetic algorithm (GA) for the simultaneous optimisation of SVM parameters and pre-processing (GENOPT-SVM). The method has been tested for the discrimination of the geographical origin of Italian olive oil (Ligurian and non-Ligurian) on the basis of near infrared (NIR) or mid infrared (FTIR) spectra. Different classification models (PLS-DA, SVM with mean centre data, GENOPT-SVM) have been tested and statistically compared using McNemars statistical test. For the two datasets, SVM with optimised pre-processing give models with higher accuracy than the one obtained with PLS-DA on pre-processed data. In the case of the NIR dataset, most of this accuracy improvement (86.3% compared with 82.8% for PLS-DA) occurred using only a single pre-processing step. For the FTIR dataset, three optimised pre-processing steps are required to obtain SVM model with significant accuracy improvement (82.2%) compared to the one obtained with PLS-DA (78.6%). Furthermore, this study demonstrates that even SVM models have to be developed on the basis of well-corrected spectral data in order to obtain higher classification rates.

Chemometric strategies to unmix information and increase the spatial description of hyperspectral images: a single-cell case study.

Hyperspectral images are analytical measurements that provide spatial and structural information. The spatial description of the samples is the specific asset of these measurements and the reason why they have become so important in (bio)chemical fields, where the microdistribution of sample constituents or the morphology or spatial pattern of sample elements constitute very relevant information. Often, because of the small size of the samples, the spatial detail provided by the image acquisition systems is insufficient. This work proposes a data processing strategy to overcome this instrumental limitation and increase the natural spatial detail present in the acquired raw images. The approach works by combining the information of a set of images, slightly shifted from each other with a motion step among them lower than the pixel size of the raw images. The data treatment includes the application of multivariate curve resolution (unmixing) multiset analysis to the set of collected images to obtain the distribution maps and spectral signatures of the sample constituents. These sets of maps are noise-filtered and compound-specific representations of all the relevant information in the pixel space and decrease the dimensionality of the original image from hundreds of spectral channels to few sets of maps, one per sample constituent or element. The information in each compound-specific set of maps is combined via a super-resolution post-processing algorithm, which takes into account the shifting, decimation, and point spread function of the instrument to reconstruct a single map per sample constituent with much higher spatial detail than that of the original image measurement.

Statistical tests for comparison of quantitative and qualitative models developed with near infrared spectral data

Near-infrared spectroscopy (NIRS) has been applied for both qualitative and quantitative evaluation of sugar beet. However, chemometrics methods are numerous and a choice criterion is sometime difficult to determine. In order to select the most accurate chemometrics method, statistical tests are developed. In the first part, quantitative models, which predict sucrose content of sugar beet, are compared. To realize a systematic study, 54 models are developed with different spectral pre-treatments (Standard Normal Variate (SNV), Detrending (D), first and second Derivative), different spectral ranges and different regression methods (Principal Component Regression (PCR), Partial Least Squares (PLS), Modified PLS (MPLS)). Analyze of variance and Fishers tests are computed to compare respectively bias and Standard Error of Prediction Corrected for bias (SEP(C)). The model developed with full spectra pre-treated by SNV, second derivative and MPLS methods gives accurate results: bias is 0.008 and SEP(C) is 0.097 g of sucrose per 100 g of sample on a concentration range between 14 and 21 g/100 g. In the second part, McNemars test is applied to compare the classification methods. The classifications are used with two data sets: the first data set concerns the disease resistance of sugar beet and the second deals with spectral differences between four spectrometers. The performances of four well-known classification methods are compared on the NIRS data: Linear Discriminant Analysis (LDA), K Nearest Neighbors method (KNN), Simple Modeling of Class Analogy (SIMCA) and Learning Vector Quantization neural network (LVQ) are computed. In this study, the most accurate method (SIMCA) has a prediction rate of 81.9% of good classification on the disease resistance determination and has 99.4% of good classification on the instrument data set.

Characterisation of heavy oils using near-infrared spectroscopy: Optimisation of pre-processing methods and variable selection

In this study, chemometric predictive models were developed from near infrared (NIR) spectra for the quantitative determination of saturates, aromatics, resins and asphaltens (SARA) in heavy petroleum products. Model optimisation was based on adequate pre-processing and/or variable selection. In addition to classical methods, the potential of a genetic algorithm (GA) optimisation, which allows the co-optimisation of pre-processing methods and variable selection, was evaluated. The prediction results obtained with the different models were compared and decision regarding their statistical significance was taken applying a randomization t-test. Finally, the results obtained for the root mean square errors of prediction (and the corresponding concentration range) expressed in %(w/w), are 1.51 (14.1-99.1) for saturates, 1.59 (0.7-61.1) for aromatics, 0.77 (0-34.5) for resins and 1.26 (0-14.7) for asphaltens. In addition, the usefulness of the proposed optimisation method for global interpretation is shown, in accordance with the known chemical composition of SARA fractions.

Monitoring polymorphic transformations by using in situ Raman hyperspectral imaging and image multiset analysis.

Polymorphism is often encountered in many crystalline compounds. To control the quality of the products, it is relevant knowing the potential presence of polymorph transformations induced by different agents, such as light exposure or temperature changes. Raman images offer a great potential to identify polymorphs involved in a process and to accurately describe this kind of solid-state transformation in the surface scanned. As a way of example, this work proposes the use of multiset analysis on the series of Raman hyperspectral images acquired during a thermal induced transformation of carbamazepine as the optimal way to extract useful information about polymorphic or any other kind of dynamic transformation among process compounds. Image multiset analysis, performed by using Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS), will furnish pure spectra and distribution maps of the compounds involved in the process and, hence, will allow the identification of polymorphs and, more important, the description of the process evolution at a global and local (pixel) level. Thus, process will be defined from a spatial point of view and by means of a set of global process profiles dependent on the process control variable. The results obtained confirm the power of this methodology and show the crucial role of the spatial information contained in the image (absent in conventional spectroscopy) for a correct process description.

Sucrose content determination of sugar beets by near infrared reflectance spectroscopy. Comparison of calibration methods and calibration transfer

The legal method (polarimetric measurement) for the determination of sucrose content since 1964 uses lead acetate. Because heavy metals are polluting, a law could forbid their use in the near future. Near infrared (NIR) spectroscopy is a suitable alternative method to replace it. For two years, 2412 samples of beet brei were analysed by NIR spectroscopy. In this article, spectral pre-processing and regression methods were compared in order to obtain an accurate prediction of sugar content. Analyse of variance and Fishers tests were calculated to compare models (bias and Standard Error of Prediction corrected for bias) in terms of statistical significance. The model developed with spectra pre-treated by standard normal variate and second derivative gave the best results. The standard error of prediction of the ratio sucrose content/fresh beet weight was low (0.11 g / 100 g). The second part of this study shows that updating of the spectral database was possible. This makes it possible to take into account new variabilities of beet. For an industrial application, calibration transfer has to be studied. Sugar beet spectra or generic standard spectra were used on two NIR instruments. A simple linear regression wavelength by wavelength gives good results for the standardisation, demonstrating a possible use of the model on different instruments. In conclusion, the replacement of the polarimetric method by NIR spectroscopy was feasible.