Automatic Classification of Laser-Induced Breakdown Spectroscopy (LIBS) Data of Protein Biomarker Solutions

Abstract

We perform multi-class classification of laser-induced breakdown spectroscopy data of four commercial samples of proteins diluted in phosphate-buffered saline solution at different concentrations: bovine serum albumin, osteopontin, leptin, and insulin-like growth factor II. We achieve this by using principal component analysis as a method for dimensionality reduction. In addition, we apply several different classification algorithms (K-nearest neighbor, classification and regression trees, neural networks, support vector machines, adaptive local hyperplane, and linear discriminant classifiers) to perform multi-class classification. We achieve classification accuracies above 98% by using the linear classifier with 21-31 principal components. We obtain the best detection performance for neural networks, support vector machines, and adaptive local hyperplanes for a range of the number of principal components with no significant differences in performance except for that of the linear classifier. With the optimal number of principal components, a simplistic K-nearest classifier still provided acceptable results. Our proposed approach demonstrates that highly accurate automatic classification of complex protein samples from laser-induced breakdown spectroscopy data can be successfully achieved using principal component analysis with a sufficiently large number of extracted features, followed by a wrapper technique to determine the optimal number of principal components.

A critical review of recent progress in analytical laser-induced breakdown spectroscopy

“…overall accuracy, robustness, or number of true and false-positive cases) should be described. Systematic studies comparing the performance of several classification methods on the same LIBS datasets are still rare [110][111][112][113], although these studies would be indispensable in assisting the selection of chemometric methods for wide use in LIBS.…”

Modeling of laser-induced breakdown spectroscopic data analysis by an automatic classifier

“…This way, assumptions of the optimal classifier can be indirectly validated on particular data. Using this approach, in [39] we demonstrated that a simplified version of the optimal classifier discussed in this study is capable of providing high classification accuracies (> 90%) when a sufficiently large number of principal components are utilized to perform multi-class classification of LIBS data of four proteins diluted in phosphate-buffered saline solution (bovine serum albumin, osteopontin, leptin, insulin-like growth factor II). This result is in agreement with the findings shown in this study that the principal components predominantly have Gaussian distribution.…”

“…Due to observed Gaussian distribution of the dark signal, this leads to conclusion that s i (λ k ) in the considered case have approximate Gaussian distribution in this range of wavelengths. Furthermore, almost all low-order principal components of the data (that are of practical importance for classification, see e.g., [39]) also have Gaussian distribution.…”

Chemometrics in laser‐induced breakdown spectroscopy

“…The fundamental idea of supervised pattern recognition is that samples with a known class as a training set are used to construct a training model and then the class or grade of an unknown sample is predicted by the training model. The common supervised pattern recognition methods mainly include partial least squares discriminate analysis (PLS‐DA), soft independent modeling of class analog (SIMCA), K‐nearest neighbor (KNN), SVM, artificial neural networks (ANN), and random forest (RF) …”

“…It calculates the nearest distance of a test sample to K of a known sample, and unknown samples can be classified by the distance of the test sample to the different types of known samples. Pokrajac et al performed multiclass classification of LIBS data for 4 commercial samples of proteins diluted in phosphate‐buffered saline solution, and PCA was used for dimension reduction. In addition, several different classification algorithms (KNN, classification and regression tree, ANN, SVM, adaptive local hyperplane, and LDA) were applied to perform multiclass classification.…”