Chemometric Strategies To Unmix Information and Increase the Spatial Description of Hyperspectral Images: A Single-Cell Case Study

Using clustering as pre-processing in the framework of signal unmixing for exhaustive exploration of archaeological artefacts in Raman imaging

Analytical chemistry on archaeological material is an essential part of modern archaeological investigations and from year to year, instrumental improvement has made it possible to generate data at a high spatial and temporal frequency. In particular, Raman spectral imaging can be successfully applied in archaeological research by its simplicity of implementation to study past human societies through the analysis of their material remains. This technique makes it possible to simultaneously obtain spatial and spectral information by preserving sample integrity. However, because of the inherent complexity of the samples in Archaeology (e.g. seniority, fragility, lack or full absence of any information about its composition), chemical interpretation can be difficult at first glance. Indeed, specific problems of spectral selectivity related to unexpected chemical compounds could appear due to their state of conservation. Furthermore, detecting minor compounds becomes challenging as major components impose their contributions in the acquired spectra. Therefore, a relevant chemometric approach has been introduced in this context to characterize distinct spectral sources in a Raman imaging dataset of an archaeological specimen - a mosaic fragment. The fragment was unearthed during the Ruscino archaeological dig on the outskirts of Perpignan, France. It dates back to the oppidum period. The aim is to extract selective spectral information from pixel clustering analysis in order to enhance the initial optimisation step within the Multivariate Curve Resolution and Alternating Least-Squares (MCR-ALS) algorithm, a well-known signal unmixing technique. The underlying principle of the MCR-ALS is that the acquired spectra can be expressed as linear combinations of pure spectra of all individual components present in the chemical system under study. Sometimes it can be difficult to obtain the desired results through the algorithm, particularly if initial estimates of spectral or concentration profiles are inaccurate due to complex signals, noise or lack of selectivity, resulting in rank deficiency (i.e. a poor estimation of the total number of pure signals). For this reason, an innovative threshold-based clustering algorithm, combined with multiple Orthogonal Projection Approaches (OPA), has been developed to improve matrix rank investigation and thus the initialisation step of the MCR-ALS approach before optimisation. The effective analysis of Raman imaging data for an archaeological mosaic played a crucial role in uncovering significant chemical information about a particular biogenic material. This insight sheds light on the origins of mortar manufacture during the oppidum period.

Handle Matrix Rank Deficiency, Noise, and Interferences in 3D Emission-Excitation Matrices: Effective Truncated Singular-Value Decomposition in Chemometrics Applied to the Analysis of Polycyclic Aromatic Compounds

The characterization of organic compounds in polluted matrices by eco-friendly three-dimensional (3D) fluorescence spectroscopy coupled with chemometric algorithms constitutes a powerful alternative to the separation techniques conventionally used. However, the systematic presence of Rayleigh and Raman scattering signals in the excitation-emission matrices (EEMs) complicates the spectral decomposition via PARAllel FACtor analysis (PARAFAC) due to the nontrilinear structure of these signals. Likewise, the specific problem of selectivity in spectroscopy for unexpected chemical components in a complex sample may render its chemical interpretation difficult at first glance. The relevant chemical information can then be complicated to extract, especially if the raw data is noisy. There are several strategies to overcome these drawbacks, but weaknesses remain. As a consequence, a new alternative method is proposed to handle these interferences, the noise, and the rank deficiencies in the data and applied for the characterization of polycyclic aromatic compound (PAC) mixtures. It is based on effective truncated singular-value decomposition (MT-SVD) that does not require any prior knowledge of the raw data. The algorithm provides a valuable estimation of the global rank to choose on complex samples where selectivity problems are observed. It is a real alternative compared to other existing methods applied to the fluorescence matrix to filter the signal from noise or light scattering effects. The first exploratory results of the proposed algorithm are promising to handle matrix rank deficiencies as well as the effects of noise and light scattering on complex PAC mixtures.

Performance assessment of linear iterative optimization technology (IOT) for Raman chemical mapping of pharmaceutical tablets

Raman chemical mapping is an inherently slow analysis tool. Accurate and robust multivariate analysis algorithms, which require least amount of time and effort in method development are desirable. Calibration-free regression and resolution approaches such as classical least squares (CLS) and multivariate curve resolution using alternating least squares (MCR-ALS), respectively, help in reducing the resources required for method development. However, conventional CLS does not consider appropriate constraints, which may result in negative and/or greater than 100 % Raman concentration scores, while MCR-ALS may not always be as accurate as regression-based algorithms. Linear iterative optimization technology (IOT) is another calibration-free algorithm, which with appropriate constraints has previously shown promise in online and offline pharmaceutical mixture composition determination. This paper aims to evaluate the performance of the linear IOT algorithm for Raman chemical mapping of the active pharmaceutical ingredient (API), diluent, and lubricant in pharmaceutical tablets. Two pre-processing strategies were applied to the raw Raman mapping spectra. The results were compared with CLS (current reference method) and MCR-ALS. Special emphasis was given to mapping at low Raman exposure times to enable feasible total acquisition times (< 5 h). The quality of IOT/CLS/MCR-ALS estimated Raman concentration predictions were assessed by calculating a correlation factor between the spectrum corresponding to the maximum predicted concentration (or resolved spectra) of a component for IOT/CLS (or MCR-ALS) and the pure powder component spectrum. The Raman chemical maps were visualized, and the average Raman concentrations scores were compared. The results demonstrated the utility of IOT in Raman chemical mapping of pharmaceutical tablets. The diluent (lactose) and API (semi-fine APAP) used in this study were reliably estimated by IOT at relatively short Raman exposure times. On the other hand, as expected, the lubricant (magnesium stearate) could not be detected in any of the cases investigated here, irrespective of the algorithm used. Overall, for the API and diluent used in this formulation as well as the chemical mapping conditions, linear IOT seemed to better estimate the pure spectrum intensities and the average Raman scores (closer to CLS) in comparison to MCR-ALS. Moreover, application of appropriate constraints in linear IOT avoided the presence of negative and/or greater than 100 % Raman concentration scores, as observed in CLS-based Raman chemical maps.

Understanding the Formation of Heartwood in Larch Using Synchrotron Infrared Imaging Combined With Multivariate Analysis and Atomic Force Microscope Infrared Spectroscopy

Formation of extractive-rich heartwood is a process in live trees that make them and the wood obtained from them more resistant to fungal degradation. Despite the importance of this natural mechanism, little is known about the deposition pathways and cellular level distribution of extractives. Here we follow heartwood formation in Larix gmelinii var. Japonica by use of synchrotron infrared images analyzed by the unmixing method Multivariate Curve Resolution - Alternating Least Squares (MCR-ALS). A subset of the specimens was also analyzed using atomic force microscopy infrared spectroscopy. The main spectral changes observed in the transition zone when going from sapwood to heartwood was a decrease in the intensity of a peak at approximately 1660 cm^-1 and an increase in a peak at approximately 1640 cm^-1. There are several possible interpretations of this observation. One possibility that is supported by the MCR-ALS unmixing is that heartwood formation in larch is a type II or Juglans-type of heartwood formation, where phenolic precursors to extractives accumulate in the sapwood rays. They are then oxidized and/or condensed in the transition zone and spread to the neighboring cells in the heartwood.

SIproc: an open-source biomedical data processing platform for large hyperspectral images

There has recently been significant interest within the vibrational spectroscopy community to apply quantitative spectroscopic imaging techniques to histology and clinical diagnosis. However, many of the proposed methods require collecting spectroscopic images that have a similar region size and resolution to the corresponding histological images. Since spectroscopic images contain significantly more spectral samples than traditional histology, the resulting data sets can approach hundreds of gigabytes to terabytes in size. This makes them difficult to store and process, and the tools available to researchers for handling large spectroscopic data sets are limited. Fundamental mathematical tools, such as MATLAB, Octave, and SciPy, are extremely powerful but require that the data be stored in fast memory. This memory limitation becomes impractical for even modestly sized histological images, which can be hundreds of gigabytes in size. In this paper, we propose an open-source toolkit designed to perform out-of-core processing of hyperspectral images. By taking advantage of graphical processing unit (GPU) computing combined with adaptive data streaming, our software alleviates common workstation memory limitations while achieving better performance than existing applications.

Tracking hidden organic carbon in rocks using chemometrics and hyperspectral imaging

Finding traces of life or organic components of prebiotic interest in the rock record is an appealing goal for numerous fields in Earth and space sciences. However, this is often hampered by the scarceness and highly heterogeneous distribution of organic compounds within rocks. We assess here an innovative analytical strategy combining Synchrotron radiation-based Fourier-Transform Infrared microspectroscopy (S-FTIR) and multivariate analysis techniques to track and characterize organic compounds at the pore level in complex oceanic rocks. S-FTIR hyperspectral images are analysed individually or as multiple image combinations (multiset analysis) using Principal Component Analyses (PCA) and Multivariate Curve Resolution – Alternating Least Squares (MCR-ALS). This approach allows extracting simultaneously pure organic and mineral spectral signatures and determining their spatial distributions and relationships. MCR-ALS analysis provides resolved S-FTIR signatures of 8 pure mineral and organic components showing the close association at a micrometric scale of organic compounds and secondary clays formed during rock alteration and known to catalyse organic synthesis. These results highlights the potential of the serpentinizing oceanic lithosphere to generate and preserve organic compounds of abiotic origin, in favour of the hydrothermal theory for the origin of life.

Identification of trace additives in polymer materials by attenuated total reflection Fourier transform infrared mapping coupled with multivariate curve resolution

Although multivariate curve resolution (MCR) has been applied to the analysis of Fourier transform infrared (FTIR) imaging, it is still problematic to determine the number of components. The reported methods at present tend to cause the components of low concentration missed. In this paper a new idea was proposed to resolve this problem. First, MCR calculation was repeated by increasing the number of components sequentially, then each retrieved pure spectrum of as-resulted MCR component was directly compared with a real-world pixel spectrum of the local high concentration in the corresponding MCR map. One component was affirmed only if the characteristic bands of the MCR component had been included in its pixel spectrum. This idea was applied to attenuated total reflection (ATR)/FTIR mapping for identifying the trace additives in blind polymer materials and satisfactory results were acquired. The successful demonstration of this novel approach opens up new possibilities for analyzing additives in polymer materials.

Multivariate Curve Resolution and Carbon Balance Constraint to Unravel FTIR Spectra from Fed-Batch Fermentation Samples

The current work investigates the capability of a tailored multivariate curve resolution-alternating least squares (MCR-ALS) algorithm to analyse glucose, phosphate, ammonium and acetate dynamics simultaneously in an E. coli BL21 fed-batch fermentation. The high-cell-density (HCDC) process is monitored by ex situ online attenuated total reflection (ATR) Fourier transform infrared (FTIR) spectroscopy and several in situ online process sensors. This approach efficiently utilises automatically generated process data to reduce the time and cost consuming reference measurement effort for multivariate calibration. To determine metabolite concentrations with accuracies between ±0.19 and ±0.96·gL^-l, the presented utilisation needs primarily-besides online sensor measurements-single FTIR measurements for each of the components of interest. The ambiguities in alternating least squares solutions for concentration estimation are reduced by the insertion of analytical process knowledge primarily in the form of elementary carbon mass balances. Thus, in this way, the established idea of mass balance constraints in MCR combines with the consistency check of measured data by carbon balances, as commonly applied in bioprocess engineering. The constraints are calculated based on online process data and theoretical assumptions. This increased calculation effort is able to replace, to a large extent, the need for manually conducted quantitative chemical analysis, leads to good estimations of concentration profiles and a better process understanding.

Confocal Raman imaging and chemometrics applied to solve forensic document examination involving crossed lines and obliteration cases by a depth profiling study

This article describes a non-destructive analytical method developed to solve forensic document examination problems involving crossed lines and obliteration.

Beyond the Sparsity-Based Target Detector: A Hybrid Sparsity and Statistics-Based Detector for Hyperspectral Images

Hyperspectral images provide great potential for target detection, however, new challenges are also introduced for hyperspectral target detection, resulting that hyperspectral target detection should be treated as a new problem and modeled differently. Many classical detectors are proposed based on the linear mixing model and the sparsity model. However, the former type of model cannot deal well with spectral variability in limited endmembers, and the latter type of model usually treats the target detection as a simple classification problem and pays less attention to the low target probability. In this case, can we find an efficient way to utilize both the high-dimension features behind hyperspectral images and the limited target information to extract small targets? This paper proposes a novel sparsity-based detector named the hybrid sparsity and statistics detector (HSSD) for target detection in hyperspectral imagery, which can effectively deal with the above two problems. The proposed algorithm designs a hypothesis-specific dictionary based on the prior hypotheses for the test pixel, which can avoid the imbalanced number of training samples for a class-specific dictionary. Then, a purification process is employed for the background training samples in order to construct an effective competition between the two hypotheses. Next, a sparse representation-based binary hypothesis model merged with additive Gaussian noise is proposed to represent the image. Finally, a generalized likelihood ratio test is performed to obtain a more robust detection decision than the reconstruction residual-based detection methods. Extensive experimental results with three hyperspectral data sets confirm that the proposed HSSD algorithm clearly outperforms the state-of-the-art target detectors.

Investigation of intervertebral disc degeneration using multivariate FTIR spectroscopic imaging

Traditionally tissue samples are analysed using protein or enzyme specific stains on serial sections to build up a picture of the distribution of components contained within them. In this study we investigated the potential of multivariate curve resolution-alternating least squares (MCR-ALS) to deconvolute 2nd derivative spectra of Fourier transform infrared (FTIR) microscopic images measured in transflectance mode of goat and human paraffin embedded intervertebral disc (IVD) tissue sections, to see if this methodology can provide analogous information to that provided by immunohistochemical stains and bioassays but from a single section. MCR-ALS analysis of non-degenerate and enzymatically in vivo degenerated goat IVDs reveals five matrix components displaying distribution maps matching histological stains for collagen, elastin and proteoglycan (PG), as well as immunohistochemical stains for collagen type I and II. Interestingly, two components exhibiting characteristic spectral and distribution profiles of proteoglycans were found, and relative component/tissue maps of these components (labelled PG1 and PG2) showed distinct distributions in non-degenerate versus mildly degenerate goat samples. MCR-ALS analysis of human IVD sections resulted in comparable spectral profiles to those observed in the goat samples, highlighting the inter species transferability of the presented methodology. Multivariate FTIR image analysis of a set of 43 goat IVD sections allowed the extraction of semi-quantitative information from component/tissue gradients taken across the IVD width of collagen type I, collagen type II, PG1 and PG2. Regional component/tissue parameters were calculated and significant correlations were found between histological grades of degeneration and PG parameters (PG1: p = 0.0003, PG2: p < 0.0001); glycosaminoglycan (GAG) content and PGs (PG1: p = 0.0055, PG2: p = 0.0001); and MRI T2* measurements and PGs (PG1: p = 0.0021, PG2: p < 0.0001). Additionally, component/tissue parameters for collagen type I and II showed significant correlations with total collagen content (p = 0.0204, p = 0.0127). In conclusion, the presented findings illustrate, that the described multivariate FTIR imaging approach affords the necessary chemical specificity to be considered an important tool in the study of IVD degeneration in goat and human IVDs.

Noise Removal with Maintained Spatial Resolution in Raman Images of Cells Exposed to Submicron Polystyrene Particles

The biodistribution of 300 nm polystyrene particles in A549 lung epithelial cells has been studied with confocal Raman spectroscopy. This is a label-free method in which particles and cells can be imaged without using dyes or fluorescent labels. The main drawback with Raman imaging is the comparatively low spatial resolution, which is aggravated in heterogeneous systems such as biological samples, which in addition often require long measurement times because of their weak Raman signal. Long measurement times may however induce laser-induced damage. In this study we use a super-resolution algorithm with Tikhonov regularization, intended to improve the image quality without demanding an increased number of collected pixels. Images of cells exposed to polystyrene particles have been acquired with two different step lengths, i.e., the distance between pixels, and compared to each other and to corresponding images treated with the super-resolution algorithm. It is shown that the resolution after application of super-resolution algorithms is not significantly improved compared to the theoretical limit for optical microscopy. However, to reduce noise and artefacts in the hyperspectral Raman images while maintaining the spatial resolution, we show that it is advantageous to use short mapping step lengths and super-resolution algorithms with appropriate regularization. The proposed methodology should be generally applicable for Raman imaging of biological samples and other photo-sensitive samples.

Quantitative fluorescence kinetic analysis of NADH and FAD in human plasma using three- and four-way calibration methods capable of providing the second-order advantage

The metabolic coenzymes reduced nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FAD) are the primary electron donor and acceptor respectively, participate in almost all biological metabolic pathways. This study develops a novel method for the quantitative kinetic analysis of the degradation reaction of NADH and the formation reaction of FAD in human plasma containing an uncalibrated interferent, by using three-way calibration based on multi-way fluorescence technique. In the three-way analysis, by using the calibration set in a static manner, we directly predicted the concentrations of both analytes in the mixture at any time after the start of their reactions, even in the presence of an uncalibrated spectral interferent and a varying background interferent. The satisfactory quantitative results indicate that the proposed method allows one to directly monitor the concentration of each analyte in the mixture as the function of time in real-time and nondestructively, instead of determining the concentration after the analytical separation. Thereafter, we fitted the first-order rate law to their concentration data throughout their reactions. Additionally, a four-way calibration procedure is developed as an alternative for highly collinear systems. The results of the four-way analysis confirmed the results of the three-way analysis and revealed that both the degradation reaction of NADH and the formation reaction of FAD in human plasma fit the first-order rate law. The proposed methods could be expected to provide promising tools for simultaneous kinetic analysis of multiple reactions in complex systems in real-time and nondestructively.

Hyperspectral image analysis. A tutorial

This tutorial aims at providing guidelines and practical tools to assist with the analysis of hyperspectral images. Topics like hyperspectral image acquisition, image pre-processing, multivariate exploratory analysis, hyperspectral image resolution, classification and final digital image processing will be exposed, and some guidelines given and discussed. Due to the broad character of current applications and the vast number of multivariate methods available, this paper has focused on an industrial chemical framework to explain, in a step-wise manner, how to develop a classification methodology to differentiate between several types of plastics by using Near infrared hyperspectral imaging and Partial Least Squares - Discriminant Analysis. Thus, the reader is guided through every single step and oriented in order to adapt those strategies to the user's case.

Pushing back the limits of Raman imaging by coupling super-resolution and chemometrics for aerosols characterization

The increasing interest in nanoscience in many research fields like physics, chemistry, and biology, including the environmental fate of the produced nano-objects, requires instrumental improvements to address the sub-micrometric analysis challenges. The originality of our approach is to use both the super-resolution concept and multivariate curve resolution (MCR-ALS) algorithm in confocal Raman imaging to surmount its instrumental limits and to characterize chemical components of atmospheric aerosols at the level of the individual particles. We demonstrate the possibility to go beyond the diffraction limit with this algorithmic approach. Indeed, the spatial resolution is improved by 65% to achieve 200 nm for the considered far-field spectrophotometer. A multivariate curve resolution method is then coupled with super-resolution in order to explore the heterogeneous structure of submicron particles for describing physical and chemical processes that may occur in the atmosphere. The proposed methodology provides new tools for sub-micron characterization of heterogeneous samples using far-field (i.e. conventional) Raman imaging spectrometer.

New chemometric approach MCR-ALS to unmix EPR spectroscopic data from complex mixtures

Electron paramagnetic resonance (EPR) spectra of mixtures are often difficult to interpret due to the superposition of spectral contribution of various species present in the complex materials. It is challenging to accurately identify the number of pure compounds present and to extract their pure spectra. In this study, the powerful chemometric method, multivariate curve resolution-alternating least squares (MCR-ALS), is applied to identify different paramagnetic centers. This method is used to simultaneously extract, with no prior knowledge, the pure spectra and the corresponding concentration profiles of all the compounds in the unknown and unresolved mixtures. The goal of our work is to apply, for the first time, this new chemometrics methodology, MCR-ALS, on EPR spectroscopic data in order to characterize a series of distinct but strongly overlapping spectra of various paramagnetic species.