Unsupervised unmixing of Raman microspectroscopic images for morphochemical analysis of non-dried brain tumor specimens

Raman microspectroscopic imaging provides molecular contrast in a label-free manner with subcellular spatial resolution. These properties might complement clinical tools for diagnosis of tissue and cells in the future. Eight Raman spectroscopic images were collected with 785 nm excitation from five non-dried brain specimens immersed in aqueous buffer. The specimens were assigned to molecular and granular layers of cerebellum, cerebrum with and without scattered tumor cells of astrocytoma WHO grade III, ependymoma WHO grade II, astrocytoma WHO grade III, and glioblastoma multiforme WHO grade IV with subnecrotic and necrotic regions. In contrast with dried tissue section, these samples were not affected by drying effects such as crystallization of lipids or denaturation of proteins and nucleic acids. The combined data sets were processed by use of the hyperspectral unmixing algorithms N-FINDR and VCA. Both unsupervised approaches calculated seven endmembers that reveal the abundance plots and spectral signatures of cholesterol, cholesterol ester, nucleic acids, carotene, proteins, lipids, and buffer. The endmembers were correlated with Raman spectra of reference materials. The focus of the single mode laser near 1 μm and the step size of 2 μm were sufficiently small to resolve morphological details, for example cholesterol ester islets and cell nuclei. The results are compared for both unmixing algorithms and with previously reported supervised spectral decomposition techniques.     

Crisp and soft multivariate methods visualize individual cell nuclei in Raman images of liver tissue sections

Raman and infrared spectroscopy have been recognized to be promising tools in clinical diagnostics because they provide molecular contrast without external stains. Here, vertex component analysis (VCA) was applied to Raman and Fourier transform infrared (FTIR) images of liver tissue sections and the results were compared with K-means cluster analysis, fuzzy C-means cluster analysis and principal component analysis. The main components of VCA from three Raman images were assigned to the central vein, periportal vein, cell nuclei, liver parenchyma and bile duct. After resonant Mie scattering correction, VCA of FTIR images identified veins, liver parenchyma, cracks, but no cell nuclei. The advantages of VCA in the context of tissue characterization by vibrational spectroscopic imaging are that the tissue architecture is visualized and the spectral information is reconstructed. Composite images were constructed that revealed a high molecular contrast and that can be interpreted in a similar way like hematoxylin and eosin stained tissue sections.     

A study of Docetaxel-induced effects in MCF-7 cells by means of Raman microspectroscopy

Chemotherapies feature a low success rate of about 25%, and therefore, the choice of the most effective cytostatic drug for the individual patient and monitoring the efficiency of an ongoing chemotherapy are important steps towards personalized therapy. Thereby, an objective method able to differentiate between treated and untreated cancer cells would be essential. In this study, we provide molecular insights into Docetaxel-induced effects in MCF-7 cells, as a model system for adenocarcinoma, by means of Raman microspectroscopy combined with powerful chemometric methods. The analysis of the Raman data is divided into two steps. In the first part, the morphology of cell organelles, e.g. the cell nucleus has been visualized by analysing the Raman spectra with k-means cluster analysis and artificial neural networks and compared to the histopathologic gold standard method hematoxylin and eosin staining. This comparison showed that Raman microscopy is capable of displaying the cell morphology; however, this is in contrast to hematoxylin and eosin staining label free and can therefore be applied potentially in vivo. Because Docetaxel is a drug acting within the cell nucleus, Raman spectra originating from the cell nucleus region were further investigated in a next step. Thereby we were able to differentiate treated from untreated MCF-7 cells and to quantify the cell–drug response by utilizing linear discriminant analysis models.     

Blind image analysis for the compositional and structural characterization of plant cell walls

A new image analysis strategy is introduced to determine the composition and the structural characteristics of plant cell walls by combining Raman microspectroscopy and unsupervised data mining methods. The proposed method consists of three main steps: spectral preprocessing, spatial clustering of the image and finally estimation of spectral profiles of pure components and their weights. Point spectra of Raman maps of cell walls were preprocessed to remove noise and fluorescence contributions and compressed with PCA. Processed spectra were then subjected to k-means clustering to identify spatial segregations in the images. Cell wall images were reconstructed with cluster identities and each cluster was represented by the average spectrum of all the pixels in the cluster. Pure components spectra were estimated by spectral entropy minimization criteria with simulated annealing optimization. Two pure spectral estimates that represent lignin and carbohydrates were recovered and their spatial distributions were calculated. Our approach partitioned the cell walls into many sublayers, based on their composition, thus enabling composition analysis at subcellular levels. It also overcame the well known problem that native lignin spectra in lignocellulosics have high spectral overlap with contributions from cellulose and hemicelluloses, thus opening up new avenues for microanalyses of monolignol composition of native lignin and carbohydrates without chemical or mechanical extraction of the cell wall materials.     

Raman microscopy for the chemometric analysis of tumor cells

Raman spectroscopy is recognized as a tool for chemometric analysis of biological materials due to the high information content relating to specific physical and chemical qualities of the sample. Thirty cells belonging to two different prostatic cell lines, PNT1A (immortalized normal prostate cell line) and LNCaP (malignant cell line derived from prostate metastases), were mapped using Raman microscopy. A range of spectral preprocessing methods (partial least-squares discriminant analyses (PLSDAs), principal component analyses (PCAs), and adjacent band ratios (ABRs)) were compared for input into linear discriminant analysis to model and classify the two cell lines. PLSDA and ABR were able to correctly classify 100% of cells into benign and malignant groups, while PLSDA correctly classified a greater proportion of individual spectra. PCA was used to image the distribution of various biochemicals inside each cell and confirm differences in composition/distribution between benign and malignant cell lines. This study has demonstrated that PLSDAs and ABRs of Raman data can identify subtle differences between benign and malignant prostatic cells in vitro.     

Are the results of customary methods for analyzing dioxin and dioxin-like compound congener profiles court-proof?

The congener profile of samples contaminated with dioxin and dioxin-like compounds allows identifying sources of contamination. This article studies the statistical methods of congener profile analysis reported in the literature with respect to the reliability of obtained results. The performance of customary analysis methods regarding raw data transformation and applied TEF (toxic equivalency factor) values is discussed. In particular, the method of principal component analysis and k-means cluster is taken as an example and examined in detail. Reasons for occurring inconsistencies such as the dependence of results on raw data transformation and the disregard of measurement uncertainty are described, and it is shown that they also explain inconsistencies in other methods of cluster analysis such as hierarchical cluster analysis and neural networks. It is concluded that these methods cannot be employed to reach court-proof decisions, i.e. decisions which meet court evidentiary standards. An alternative approach to analyzing congener profiles based on mathematical statistics is briefly presented, allowing reliable, court-proof decisions.     

Aerosol time-of-flight mass spectrometry data analysis: a benchmark of clustering algorithms

Airborne particulate matter is an important component of atmospheric pollution, affecting human health, climate, and visibility. Modern instruments allow single particles to be analyzed one-by-one in real time, and offer the promise of determining the sources of individual particles based on their mass spectral signatures. The large number of particles to be apportioned makes clustering a necessary step. The goal of this study is to compare using mass spectral data the accuracy and speed of several clustering algorithms: ART-2a, several variants of hierarchical clustering, and K-means. Repeated simulations with various algorithms and different levels of data preprocessing suggest that hierarchical clustering methods using derivatives of Ward's algorithm discriminate sources with fewer errors than ART-2a, which itself discriminates much better than point-wise hierarchical clustering methods. In most cases, K-means algorithms do almost as well as the best hierarchical clustering. These efficient algorithms (clustering derived from Ward's algorithm, ART-2a and K-means) are most accurate when the relative peak areas have been pre-scaled by taking the square root. Analysis times vary within a factor of 30, and when accuracy above 95% is required, run times scale up as the square of the number of particles. Algorithms derived from Ward's remain the most accurate under a wide range of conditions and conversely, for an equal accuracy, can deliver a shorter list of clusters, allowing faster and maybe on-the-fly classification.     

Kernel principal component analysis residual diagnosis (KPCARD): An automated method for cosmic ray artifact removal in Raman spectra

A new, fully automated, rapid method, referred to as kernel principal component analysis residual diagnosis (KPCARD), is proposed for removing cosmic ray artifacts (CRAs) in Raman spectra, and in particular for large Raman imaging datasets. KPCARD identifies CRAs via a statistical analysis of the residuals obtained at each wavenumber in the spectra. The method utilizes the stochastic nature of CRAs; therefore, the most significant components in principal component analysis (PCA) of large numbers of Raman spectra should not contain any CRAs. The process worked by first implementing kernel PCA (kPCA) on all the Raman mapping data and second accurately estimating the inter- and intra-spectrum noise to generate two threshold values. CRA identification was then achieved by using the threshold values to evaluate the residuals for each spectrum and assess if a CRA was present.     

Multivariate classification of pigments and inks using combined Raman spectroscopy and LIBS

The authenticity of objects and artifacts is often the focus of forensic analytic chemistry. In document fraud cases, the most important objective is to determine the origin of a particular ink. Here, we introduce a new approach which utilizes the combination of two analytical methods, namely Raman spectroscopy and laser-induced breakdown spectroscopy (LIBS). The methods provide complementary information on both molecular and elemental composition of samples. The potential of this hyphenation of spectroscopic methods is demonstrated for ten blue and black ink samples on white paper. LIBS and Raman spectra from different inks were fused into a single data matrix, and the number of different groups of inks was determined through multivariate analysis, i.e., principal component analysis, soft independent modelling of class analogy, partial least-squares discriminant analysis, and support vector machine. In all cases, the results obtained with the combined LIBS and Raman spectra were found to be superior to those obtained with the individual Raman or LIBS data sets.     

Selecting variables for k-means cluster analysis by using a genetic algorithm that optimises the silhouettes

The goal of present work is to analyse the effect of having non-informative variables (NIV) in a data set when applying cluster analysis and to propose a method computationally capable of detecting and removing these variables. The method proposed is based on the use of a genetic algorithm to select those variables important to make the presence of groups in data clear. The procedure has been implemented to be used with k-means and using the cluster silhouettes as fitness function for the genetic algorithm.The main problem that can appear when applying the method to real data is the fact that, in general, we do not know a priori what the real cluster structure is (number and composition of the groups).The work explores the evolution of the silhouette values computed from the clusters built by using k-means when non-informative variables are added to the original data set in both a literature data set as well as some simulated data in higher dimension. The procedure has also been applied to real data sets.     

Raman spectral imaging of single cancer cells: probing the impact of sample fixation methods

Raman spectroscopy has proven its potential for the analysis of cell constituents and processes. However, sample preparation methods compatible with clinical practice must be implemented for collection of accurate spectral information. This study aims at assessing, using micro-Raman imaging, the effects of some routinely used fixation methods such as formalin-fixation, formalin-fixation/air drying, cytocentrifugation, and air drying on intracellular spectral information. Data were compared with those acquired from single living cells. In parallel to these spectral information, cell morphological modifications that accompany sample preparation were compared. Spectral images of isolated cells were first analyzed in an unsupervised way using hierarchical cluster analysis (HCA), which allowed delimitation of the cellular compartments. The resulting nuclei cluster centers were compared and revealed at the molecular level that fixation induced changes in spectral information assigned to nucleic acids and proteins. In a second approach, a supervised fitting procedure using model spectra of DNA, RNA, and proteins, chemically extracted from living cells, revealed very small modifications at the level of the localization and quantification of these macromolecules. Finally, HCA and principal components analysis (PCA) performed on individual spectra randomly selected from the nuclear regions showed that formalin-fixation and cytocentrifugation are sample preparation methods that have little impact on the biochemical information as compared to living conditions. Any step involving cell air drying seems to accentuate the spectral deviations from the other preparation methods. It is therefore important in a future context of spectral cytology to take into account these variations.     

自适应拉曼光谱成像数据去噪及其在植物细胞壁光谱分析中的应用

拉曼光谱成像数据存在基线漂移与宇宙射线干扰峰两类噪声信号,无法直接用于光谱分析研究,必须去除。现有单光谱去噪方法处理结果不稳定、可重复性差。针对这一问题,本研究提出了一种自适应拉曼光谱成像数据新型去噪法,采用优化的自适应迭代惩罚最小二乘法( Adaptive iteratively reweighted penalized least-squares,airPLS)和基于主成分分析( PCA)的干扰峰消除算法修正光谱基线漂移和宇宙射线干扰峰,具有输入参数少、光谱失真小、处理速度快、去噪结果稳定等优点。利用本方法去除了芒草( Miscanthus sinensis)细胞壁拉曼光谱成像数据(9010条光谱)中的噪声信号,并对去噪后数据进行PCA和聚类分析(CA),成功区分非植物光谱与植物光谱,分类结果优于未去噪数据。预期本方法可应用于其它光谱成像数据去噪,为光谱的解译和定量分析提供可靠的研究基础。     

Testing the performance of pure spectrum resolution from Raman hyperspectral images of differently manufactured pharmaceutical tablets

Chemical imaging is a rapidly emerging analytical method in pharmaceutical technology. Due to the numerous chemometric solutions available, characterization of pharmaceutical samples with unknown components present has also become possible. This study compares the performance of current state-of-the-art curve resolution methods (multivariate curve resolution-alternating least squares, positive matrix factorization, simplex identification via split augmented Lagrangian and self-modelling mixture analysis) in the estimation of pure component spectra from Raman maps of differently manufactured pharmaceutical tablets. The batches of different technologies differ in the homogeneity level of the active ingredient, thus, the curve resolution methods are tested under different conditions. An empirical approach is shown to determine the number of components present in a sample. The chemometric algorithms are compared regarding the number of detected components, the quality of the resolved spectra and the accuracy of scores (spectral concentrations) compared to those calculated with classical least squares, using the true pure component (reference) spectra. It is demonstrated that using appropriate multivariate methods, Raman chemical imaging can be a useful tool in the non-invasive characterization of unknown (e.g. illegal or counterfeit) pharmaceutical products.     

Using probabilistic relational learning to support bronchial carcinoma diagnosis based on ion mobility spectrometry

Ion Mobility Spectrometry (IMS) provides a means for analyzing the substances a person exhales. In this paper, we report on an approach to support early diagnosis of bronchial carcinoma based on such IMS measurements. Given the peaks in a set of ion mobility spectra, we first cluster these peaks with a modified k-means algorithm. We then apply probabilistic relational modelling and learning methods to a logical representation of the data obtained from the ion mobility spectra and the peak clusters. Markov Logic Networks and the MLN system Alchemy are employed for various modelling and learning scenarios. These scenarios are evaluated with respect to ease of use, classification accuracy, and knowledge representation aspects.     

Fast detection of high-order epistatic interactions in genome-wide association studies using information theoretic measure

There are many algorithms for detecting epistatic interactions in GWAS. However, most of these algorithms are applicable only for detecting two-locus interactions. Some algorithms are designed to detect only two-locus interactions from the beginning. Others do not have limits to the order of interactions, but in practice take very long time to detect higher order interactions in real data of GWAS. Even the better ones take days to detect higher order interactions in WTCCC data.We propose a fast algorithm for detection of high order epistatic interactions in GWAS. It runs k-means clustering algorithm on the set of all SNPs. Then candidates are selected from each cluster. These candidates are examined to find the causative SNPs of k-locus interactions. We use mutual information from information theory as the measure of association between genotypes and phenotypes.We tested the power and speed of our method on extensive sets of simulated data. The results show that our method has more or equal power, and runs much faster than previously reported methods. We also applied our algorithm on each of seven diseases in WTCCC data to analyze up to 5-locus interactions. It takes only a few hours to analyze 5-locus interactions in one dataset. From the results we make some interesting and meaningful observations on each disease in WTCCC data.In this study, a simple yet powerful two-step approach is proposed for fast detection of high order epistatic interaction. Our algorithm makes it possible to detect high order epistatic interactions in GWAS in a matter of hours on a PC.     

Raman spectroscopy in combination with background near-infrared autofluorescence enhances the in vivo assessment of malignant tissues

The diagnostic ability of optical spectroscopy techniques, including near-infrared (NIR) Raman spectroscopy, NIR autofluorescence spectroscopy and the composite Raman and NIR autofluorescence spectroscopy, for in vivo detection of malignant tumors was evaluated in this study. A murine tumor model, in which BALB/c mice were implanted with Meth-A fibrosarcoma cells into the subcutaneous region of the lower back, was used for this purpose. A rapid-acquisition dispersive-type NIR Raman system was employed for tissue Raman and NIR autofluorescence spectroscopic measurements at 785-nm laser excitation. High-quality in vivo NIR Raman spectra associated with an autofluorescence background from mouse skin and tumor tissue were acquired in 5 s. Multivariate statistical techniques, including principal component analysis (PCA) and linear discriminant analysis (LDA), were used to develop diagnostic algorithms for differentiating tumors from normal tissue based on their spectral features. Spectral classification of tumor tissue was tested using a leave-one-out, cross-validation method, and the receiver operating characteristic (ROC) curves were used to further evaluate the performance of diagnostic algorithms derived. Thirty-two in vivo Raman, NIR fluorescence and composite Raman and NIR fluorescence spectra were analyzed (16 normal, 16 tumors). Classification results obtained from cross-validation of the LDA model based on the three spectral data sets showed diagnostic sensitivities of 81.3%, 93.8% and 93.8%; specificities of 100%, 87.5% and 100%; and overall diagnostic accuracies of 90.6%, 90.6% and 96.9% respectively, for tumor identification. ROC curves showed that the most effective diagnostic algorithms were from the composite Raman and NIR autofluorescence techniques.     

Near-Infrared Raman Spectroscopy for In Vitro Detection of Cervical Precancers

Abstract— In this study, we investigate the potential of near-infrared Raman spectroscopy to differentiate cervical precancers from normal tissues, inflammation and metaplasia and to differentially diagnose low-grade and high-grade precancers. Near infrared Raman spectra were measured from 36 biopsies from 18 patients in vitro. Detection algorithms were developed and evaluated relative to histopathologic examination. Algorithms based on empirically selected peak intensities, ratios of peak intensities and a combination of principal component analysis for data reduction and Fisher discriminant analysis for classification were investigated. Spectral peaks were tentatively identified from measured spectra of potential chromophores. Empirically selected normalized intensities can differentiate precancers from other tissues with an average sensitivity and specificity of 88 ± 4% and 92 ± 4%. Ratios of un-normalized intensities can differentiate precancers from other tissues with a sensitivity and specificity of 82% and 88% and high-grade from low-grade lesions with a sensitivity and specificity of 100%. Using multivariate methods, intensities at eight frequencies can be used to differentiate precancers from all other tissues with a sensitivity and specificity of 82% and 92% in an unbiased test. Raman algorithms can potentially separate benign abnormalities such as inflammation and metaplasia from precancers. Comparison of tissue spectra to published and measured chromophore spectra indicate that the most likely primary contributors to the tissue spectra are collagen, nucleic acids, phospholipids and glucose 1-phos-phate. These results suggest that near-infrared Raman spectroscopy can be used for cervical precancer diagnosis and may be able to accurately separate samples with inflammation and metaplasia from precancer.     

Classification of nasopharyngeal cell lines (C666-1, CNE2, NP69) via Raman spectroscopy and decision tree

Researchers have demonstrated that Raman spectroscopy can be used for characterization of tumor cells with excellent spatial resolution. However, performance evaluation of different algorithms in classifying multiclass of Raman spectra has not been reported yet. In this work, we present Raman spectra of nasopharyngeal carcinoma and nasopharyngeal normal cell lines. Combined with student’s t-test and several multivariate approaches, including decision tree, support vector classification, and linear discriminant analysis, our work shows that the relative content of two histological abnormality sensitive bands at 1449 and 1658 cm⁻¹ in tumor cells is significantly different from that of normal cells (p = 0.0132), and can be a biomarker to classify these cells. This difference is confirmed by importance analyses in the decision tree model. Furthermore, performances of statistical methods are compared with one another to explore the ability in classification. Results show that the decision tree can be more capable for classification between tumorous and normal cell lines with sensitivity and specificity of 99.0% and 96.9%, respectively. Findings of this work further support our previous work and indicate that the decision tree performs more robustly in cell classification. Our work will prove helpful to the early diagnosis of nasopharyngeal carcinoma, and will indicate the decision tree to be the primary algorithm in tumor-cell classification.     

Discrimination of natural gas-related bacteria by means of micro-Raman spectroscopy

Methane-oxidizing bacteria (MOB) are a unique group of gram-negative bacteria that are proved to be biological indicator for gas prospecting since they utilize methane as a sole source of carbon and energy. Herein the feasibility of a novel and efficient gas prospecting method using Raman spectroscopy is studied. Confocal Raman spectroscopy is utilized to establish a Raman database of 11 species of methanotrophs and other closely related bacteria with similar morphology that generally coexist in the upper soil of natural gas. After strict and consistent spectral preprocessing, Raman spectra from the whole cell area are analyzed using the combination of principal component analysis (PCA) and Mahalanobis distance (MD) that allow unambiguous classification of the different cell types with an accuracy of 95.91%. The discrimination model based on multivariate analysis is further evaluated by classifying Raman spectra from independently cultivated bacteria, and achieves an overall accuracy of 94.04% on species level. Our approach using Raman spectroscopy in combination with statistical analysis of various gas reservoirs related bacteria provides rapid distinction that can potentially play a vital role in gas exploration.     

Characterization and analysis of mycobacteria and Gram-negative bacteria and co-culture mixtures by Raman microspectroscopy, FTIR, and atomic force microscopy

The molecular composition of mycobacteria and Gram-negative bacteria cell walls is structurally different. In this work, Raman microspectroscopy was applied to discriminate mycobacteria and Gram-negative bacteria by assessing specific characteristic spectral features. Analysis of Raman spectra indicated that mycobacteria and Gram-negative bacteria exhibit different spectral patterns under our experimental conditions due to their different biochemical components. Fourier transform infrared (FTIR) spectroscopy, as a supplementary vibrational spectroscopy, was also applied to analyze the biochemical composition of the representative bacterial strains. As for co-cultured bacterial mixtures, the distribution of individual cell types was obtained by quantitative analysis of Raman and FTIR spectral images and the spectral contribution from each cell type was distinguished by direct classical least squares analysis. Coupled atomic force microscopy (AFM) and Raman microspectroscopy realized simultaneous measurements of topography and spectral images for the same sampled surface. This work demonstrated the feasibility of utilizing a combined Raman microspectroscopy, FTIR, and AFM techniques to effectively characterize spectroscopic fingerprints from bacterial Gram types and mixtures.