Merge pull request #191 from chhoumann/kb-270-new-cite-pca

[KB-270] Updated PCA sources

report_thesis/src/references.bib

@@ -794,3 +794,57 @@ @article{el_haddad_ann_2013
 	keywords = {Artificial neural network, Laser-induced breakdown spectroscopy (LIBS), Quantitative analysis, Soil},
 	pages = {51--57},
 }
+
+@article{pca_review_paper,
+title = {On the utilization of principal component analysis in laser-induced breakdown spectroscopy data analysis, a review},
+journal = {Spectrochimica Acta Part B: Atomic Spectroscopy},
+volume = {148},
+pages = {65-82},
+year = {2018},
+issn = {0584-8547},
+doi = {https://doi.org/10.1016/j.sab.2018.05.030},
+url = {https://www.sciencedirect.com/science/article/pii/S0584854718301526},
+author = {Pavel Pořízka and Jakub Klus and Erik Képeš and David Prochazka and David W. Hahn and Jozef Kaiser},
+abstract = {An implementation of a fast, robust, and effective algorithm is inevitable in modern multivariate data analysis (MVDA). The principal component analysis (PCA) algorithm is becoming popular not only in the spectroscopic community because it complies with the qualities mentioned above. PCA is, therefore, often used for the processing of detected multivariate signal (characteristic spectra). Over the past decade, PCA has been adopted by the Laser-Induced Breakdown Spectroscopy (LIBS) community and the number of scientific articles referring to PCA steadily increases. The interest in PCA is not caused only by the basic need to obtain a fast data visualization on a lower dimensional scale and to inspect the most prominent variables. Most recently, PCA has also been applied to yield unconventional data analyses, i.e. processing of large scale LIBS maps. However, a rapid development of LIBS-related instrumentation and applications has led to some non-uniform methodologies in the implementation and utilization of MVDA, including PCA. Thus, in this work, we critically assess and elaborate on the approaches to utilize PCA in LIBS data processing. The aim of this article is also to derive some implications and to suggest advice in data preprocessing, visualization, dimensionality reduction, model building, classification, quantification and non-conventional multivariate mapping. This review reflects also other MVDA algorithms than PCA and consequently, presented conclusions and recommendations can be generalized.}
+}
+
+@Article{moncayo_pca,
+author ="Moncayo, Samuel and Duponchel, Ludovic and Mousavipak, Niloofar and Panczer, Gérard and Trichard, Florian and Bousquet, Bruno and Pelascini, Frédéric and Motto-Ros, Vincent",
+title  ="Exploration of megapixel hyperspectral LIBS images using principal component analysis",
+journal  ="J. Anal. At. Spectrom.",
+year  ="2018",
+volume  ="33",
+issue  ="2",
+pages  ="210-220",
+publisher  ="The Royal Society of Chemistry",
+doi  ="10.1039/C7JA00398F",
+url  ="http://dx.doi.org/10.1039/C7JA00398F",
+abstract  ="Laser-Induced Breakdown Spectroscopy (LIBS) has achieved promising performance as an elemental imaging technology{,} and considerable progress has been achieved in the development of LIBS over the last several years{,} which has led to great interest in the use of LIBS in various fields of applications. LIBS is a highly attractive technology that is distinguished by its table top instrumentation{,} speed of operation{,} and operation in ambient atmosphere{,} able to produce megapixel multi-elemental images with micrometric resolution (10 μm) and ppm-scale sensitivity. However{,} the points that limit the development of LIBS are undeniably the expertise and the time required to extract a relevant signal from the LIBS dataset. The complexity of the emission spectra (e.g.{,} elemental responses{,} structure of the baseline){,} the high dynamic range of measurement (i.e.{,} possibility to image major to trace elements){,} and the large number of spectra to process require new data analysis strategies. Such new strategies are particularly critical for multi-phase materials. In this paper{,} we report a new methodology based on the well-known Principal Component Analysis (PCA) approach for the multivariate hyperspectral analysis of LIBS images. The proposed methodology is designed for large{,} raw{,} and potentially complex series of LIBS spectra{,} that allows various and exhaustive levels of information to be extracted (including the characterization of mineral phases{,} assessment of the measurement and identification of isolated elements) and facilitates the manipulation of such hyperspectral datasets."}
+
+@article{porizka_pca,
+title = {Laser-induced breakdown spectroscopy for in situ qualitative and quantitative analysis of mineral ores},
+journal = {Spectrochimica Acta Part B: Atomic Spectroscopy},
+volume = {101},
+pages = {155-163},
+year = {2014},
+issn = {0584-8547},
+doi = {https://doi.org/10.1016/j.sab.2014.08.027},
+url = {https://www.sciencedirect.com/science/article/pii/S058485471400202X},
+author = {P. Pořízka and A. Demidov and J. Kaiser and J. Keivanian and I. Gornushkin and U. Panne and J. Riedel},
+keywords = {Laser-induced breakdown spectroscopy, LIBS, Chemometrics, Principal component analysis, Geochemical analysis},
+abstract = {In this work, the potential of laser-induced breakdown spectroscopy (LIBS) for discrimination and analysis of geological materials was examined. The research was focused on classification of mineral ores using their LIBS spectra prior to quantitative determination of copper. Quantitative analysis is not a trivial task in LIBS measurement because intensities of emission lines in laser-induced plasmas (LIP) are strongly affected by the sample matrix (matrix effect). To circumvent this effect, typically matrix-matched standards are used to obtain matrix-dependent calibration curves. If the sample set consists of a mixture of different matrices, even in this approach, the corresponding matrix has to be known prior to the downstream data analysis. For this categorization, the multielemental character of LIBS spectra can be of help. In this contribution, a principal component analysis (PCA) was employed on the measured data set to discriminate individual rocks as individual matrices against each other according to their overall elemental composition. Twenty-seven igneous rock samples were analyzed in the form of fine dust, classified and subsequently quantitatively analyzed. Two different LIBS setups in two laboratories were used to prove the reproducibility of classification and quantification. A superposition of partial calibration plots constructed from the individual clustered data displayed a large improvement in precision and accuracy compared to the calibration plot constructed from all ore samples. The classification of mineral samples with complex matrices can thus be recommended prior to LIBS system calibration and quantitative analysis.}
+}
+
+@article{sirven_pca_ann_plsr,
+  author = {Sirven, J.-B. and Bousquet, B. and Canioni, L. and Sarger, L.},
+  title = {Laser-Induced Breakdown Spectroscopy of Composite Samples: Comparison of Advanced Chemometrics Methods},
+  journal = {Analytical Chemistry},
+  volume = {78},
+  number = {5},
+  pages = {1462-1469},
+  year = {2006},
+  doi = {10.1021/ac051721p},
+  note = {PMID: 16503595},
+  url = {https://doi.org/10.1021/ac051721p},
+  eprint = {https://doi.org/10.1021/ac051721p}
+}

report_thesis/src/sections/related_work.tex

@@ -50,9 +50,18 @@ \subsection{Preprocessing and Feature Engineering}
 \subsection{Dimensionality Reduction Techniques}
 Dimensionality reduction techniques such as \gls{pca} play a crucial role in managing the high-dimensional nature of \gls{libs} data.
 
-\citet{rezaei_dimensionality_reduction} explored various machine learning techniques combined with \gls{pca} for dimensionality reduction. Their results demonstrated that \gls{svr} with \gls{pca} performed the best for different elements, highlighting the effectiveness of dimensionality reduction techniques in improving model performance by managing the complexities of high-dimensional data.
+\citet{pca_review_paper} conducted a comprehensive review of \gls{pca} applied within the context of \gls{libs}.
+This review highlighted numerous studies that successfully utilized \gls{pca}.
+For instance, \citet{moncayo_pca} used \gls{pca} to analyze megapixel elemental maps composed of over one million \gls{libs} spectra.
+The \gls{pca} approach effectively separated the contributions of various minerals, including those present in low concentrations, demonstrating its robustness in handling highly diverse samples.
 
-The study by \citet{liuComparisonQuantitativeAnalysis2022} explores the use of \gls{marscode} \gls{libs} for quantitative analysis of olivine in a simulated Martian atmosphere, focusing on multivariate analysis methods to address challenges posed by \gls{libs} data, such as high dimensionality and multicollinearity.
-The methods evaluated include \gls{ulr}, \gls{mvlr}, \gls{pcr}, \gls{plsr}, ridge regression, \gls{lasso}, \gls{enet}, and \gls{bpnn}.
-The findings demonstrate the effectiveness of dimension reduction techniques, especially \gls{plsr}, and nonlinear analysis for improving quantitative analysis accuracy of olivine using \gls{libs} data.
-This approach is particularly relevant to our work due to the focus on advanced statistical methods and machine learning algorithms for handling complex, high-dimensional \gls{libs} data, aligning with our objectives of improving accuracy and robustness in predicting major oxide compositions.
+In another example, \citet{porizka_pca} employed \gls{pca} to filter outliers and classify samples based on their matrix composition, including elements such as \ce{Al}, \ce{Ca}, \ce{Na}, and \ce{Si}. 
+This classification was followed by a univariate calibration of copper (\ce{Cu}) in soil samples, resulting in reduced bias.
+Additionally, \gls{pca} was used to discriminate individual rocks based on their overall elemental composition, effectively addressing matrix effects that can significantly impact the accuracy of analytical results.
+These studies, among others highlighted by \cite{pca_review_paper}, underscore the effectiveness of \gls{pca} as a preprocessing technique in \gls{libs} analysis.
+
+\citet{sirven_pca_ann_plsr} investigated the influence of matrix effects on the performance of quantitative analysis of chromium (\ce{Cr}) in soil samples using \gls{libs}.
+\gls{pca} was used to classify spectra from two different soils and to detect outliers.
+It successfully separated spectra from agricultural soil and kaolinite in the plane of the first two components.
+Furthermore, \gls{pca} was used to identify and remove outliers from the dataset, enhancing the accuracy of subsequent analyses using \gls{ann}s and \gls{plsr}.
+This study demonstrated the utility of \gls{pca} in managing matrix effects and improving the accuracy of quantitative \gls{libs} analysis.