Improvement of pixel classification by the simultaneous use of spectral and spatial information in the framework of spectroscopic imaging.

Hyperspectral imaging technology is developing in a very fast way. We find it today in many analytical developments using different spectroscopies for sample classification purposes. Instrumental developments allow us to acquire more and more data in shorter and shorter periods of time while improving their quality. Therefore, we are going in the right direction as far as the measure is concerned. On the other hand, we can make a more mixed assessment for the hyperspectral imaging data processing. Indeed, the data acquired in spectroscopic imaging have the particularity of encoding both spectral and spatial information. Unfortunately, in chemometrics, almost all classification approaches today only use spectral information from three-dimensional hyperspectral data arrays. To be more precise, an approach encompassing the unfolding/refolding of such arrays is often applied beforehand because the majority of algorithms for analysing these data are not capable of handling them in their original structure. Spatial information is therefore lost during the chemometric exploration. The study of the spectral part of the acquired data array alone is clearly a limitation that we propose to overcome in this work. 2-D Stationary Wavelet Transform will be used in the data preprocessing phase to ensure the joint use of spectral and spatial information. Two spectroscopic datasets will then be used to evaluate the potential of our approach in the context of supervised classification.

Data fusion of LIBS and PIL hyperspectral imaging: Understanding the luminescence phenomenon of a complex mineral sample.

Laser-induced breakdown spectroscopy (LIBS) imaging is an innovative technique that associates the valuable atomic, ionic and molecular emission signals of the parent spectroscopy with spatial information. LIBS works using a powerful pulse laser as excitation source, to generate a plasma exhibiting emission lines of atoms, ions and molecules present in the ablated matter. The advantages of LIBS imaging are potential high sensitivity (in the order of ppm), easy sample preparation, fast acquisition rate (up to 1 kHz) and µm scale spatial resolution (weight of the ablated material in the order of ng). Despite these positive aspects, LIBS imaging easily provides datasets consisting of several million spectra, each containing several thousand spectral channels. Under these conditions, the current chemometric analyses of the raw data are still possible, but require too high computing resources. Therefore, the aim of this work is to propose a data compression strategy oriented to keep the most relevant spectral channel and pixel information to facilitate, fast and reliable signal unmixing for an exhaustive exploration of complex samples. This strategy will apply not only to the context of LIBS image analysis, but to the fusion of LIBS with other imaging technologies, a scenario where the data compression step becomes even more mandatory. The data fusion strategy will be applied to the analysis of a heterogeneous kyanite mineral sample containing several trace elements by LIBS imaging associated with plasma induced luminescence (PIL) imaging, these two signals being acquired simultaneously by the same microscope. The association of compression and spectral data fusion will allow extracting the compounds in the mineral sample associated with a fused LIBS/PIL fingerprint. This LIBS/PIL association will be essential to interpret the PIL spectral information, which is nowadays very complex due to the natural overlapped signals provided by this technique.

Saturated signals in spectroscopic imaging: why and how should we deal with this regularly observed phenomenon?

We have all been confronted one day by saturated signals observed on acquired spectra, whatever the technique considered. A saturation, also known as clipping in signal processing, is a form of distortion that limits a signal once it exceeds a threshold. As a consequence, clipped or saturated bands with their characteristic plateau present numerical values that do not correspond to the analytical reality of the analyzed sample. Of course, analysts know that they cannot consider these erroneous values and therefore reconsider either sample preparation or instrument settings. Unfortunately, there are many experiments today (and this is the case in spectroscopic imaging) for which we will not be able to fight against the saturation effect that will undeniably be observed on the acquired spectra. The aim of this article is first to show why it is important to correct these saturation effects at the risk of having a biased view of the sample and more specifically in the context of multivariate data analysis. In a second step, we will look at strategies for managing saturated bands. An original concept will then be presented by considering saturated values as missing ones. A statistical imputation strategy will then be implemented in order to recover the information lost during the measurement.

Fusing spectral and spatial information with 2-D stationary wavelet transform (SWT 2-D) for a deeper exploration of spectroscopic images.

Nowadays, it is clear that there is an increasing importance in spectroscopic imaging in all fields of science. Obviously, one bulk analysis can no longer be satisfactory, as the interest focuses more on the chemical nature and the location of the compounds present within a given complex matrix. This is, evidently, due to the fact that for a more comprehensive exploration of complex samples, one single acquired hyperspectral data cube can provide both spectral and spatial information simultaneously. Although many techniques were proposed by the chemometric community in explorations of these specific datasets, unfortunately, they are almost always focusing on spectral information, even if chemical images were ultimately observed. In other words, spatial information is not well exploited, and therefore lost during the actual chemometric calculation phase. The goal of this short communication is to present a very simple and fast spectral/spatial fusion approach based on 2-D stationary wavelet transform (SWT 2-D) which is able to improve the obtainable information, compared with a classical data analysis, in which the spatial domain would not be considered nor used.

Randomised SIMPLISMA: Using a dictionary of initial estimates for spectral unmixing in the framework of chemical imaging.

Hyperspectral imaging opens the opportunity in analytical chemistry to investigate always more complex samples by the use of Multivariate Curve Resolution - Alternating Least Squares (MCR-ALS) and other signal unmixing techniques, but not without difficulties. Nowadays, one of the principal challenges regarding this kind of analysis is the awkward estimation of the correct chemical rank of the dataset, which represents the total number of pure compounds present in the chemical system. Despite the existence of various algorithms able to focus on this rank evaluation, the method very often used for this task is finally quite simple since it is based on the observation of the eigenvalues generated by the Principal Component Analysis (PCA). Although this method has shown some potential for rank evaluation, it is still difficult to use it on complex and big datasets or when the signal to noise ratio is relatively weak. In this paper, we introduce a new method, based on the SIMPLE-to-use Self-modeling Mixture Analysis (SIMPLISMA) algorithm that we call Randomised SIMPLISMA. The main idea is thus to use random selections of spectra from the initial dataset and to apply the SIMPLISMA approach to each of them. At the end of this step, all selected spectra are observed using PCA where observed clusters can potentially be highlighted and exploited for the tasks we are interested in. With the present paper, we want to highlight in particular the possibility of an easier rank estimation and initial estimates generation when this approach is considered. Datasets of different complexity acquired with various spectroscopic techniques will be explored in order to evaluate the potential of this approach.

Detection of minor compounds in complex mineral samples from millions of spectra: A new data analysis strategy in LIBS imaging.

Today, Laser-Induced Breakdown Spectroscopy (LIBS) imaging is in full change. Indeed, always more stable instrumentations are developed, which significantly increases the signal quality and naturally the analytical potential of the technique for the characterization of complex and heterogeneous samples at the micro-scale level. Obviously, other intrinsic features such as a limit of detection in the order of ppm, a high field of view and high acquisition rate make it one of the most complete chemical imaging techniques to date. It is thus possible in these conditions to acquire several million spectra from one single sample in just hours. Managing big data in LIBS imaging is the challenge ahead. In this paper, we put forward a new spectral analysis strategy, called embedded k-means clustering, for simultaneous detection of major and minor compounds and the generation of associated localization maps. A complex rock section with different phases and traces will be explored to demonstrate the value of this approach.

Multi-block classification of Italian semolina based on Near Infrared Spectroscopy (NIR) analysis and alveographic indices.

Durum wheat (Triticum turgidum ssp. durum) is widely grown in the Mediterranean area. The semolina obtained by this grain is used to prepare pasta, couscous, and baked products all over the world. The growing area affects the characteristics of Durum wheat; consequently, it is relevant to trace this product. The present study aims at developing an analytical methodology which would allow tracing durum semolina harvested in 7 different Italian macro-areas. In order to achieve this goal, 597 samples of semolina have been analysed by Near Infrared Spectroscopy, and by measuring alveographic parameters. Eventually, the information collected have been handled by a multi-block classifier (SO-PLS-LDA) in order to predict the origin of samples. The proposed approach provided extremely satisfactory results (in external validation, on a test set of 140 objects), correctly classifying all samples according to their growing area, confirming it represents a suitable solution for tracing durum wheat semolina.