LAboratoire de Spectrochimie Infrarouge et Raman – UMR 8516
  • Français
  • English
  • Spanish

Description of the subject

Chemometric exploration in hyperspectral imaging facing multimodality and big data.
Supervisor(s): L. Duponchel

Analytical chemistry is changing rapidly. Indeed we acquire always more data in order to go ever further in the exploration of our complex samples. This stronger growth is mainly explained by the development of new instrumental concepts and new experimental methodologies. It is obvious that the nature and the structure of acquired data change significantly. Hyperspectral imaging has not escaped this trend. It quickly became tool of choice for molecular characterisation of complex samples in many scientific domains. The main reason is that it simultaneously provides spectral and spatial information. As a result, chemometrics has provided many exploration tools (PCA, clustering, MCR-ALS …) well-suited for such data structure at early stage. However we are today facing a new challenge considering multimodality which will be present everywhere in the next ten years. In our case of hyperspectral imaging, it corresponds to the analysis of the same sample region of interest explored with different spectroscopic techniques. It is obvious that chemometrics already provides tools in order to analyse separately each acquired hyperspectral data cube. In this thesis, we aim to develop new methods around the concept of data fusion for a simultaneous analysis of all modalities allowing us to go deeper in the characterisation of complex samples. In order to meet this requirement, these are questions that we still have to answer. How is it possible to really analyse the same region of interest in a sample for all considered spectroscopy at the micron scale? If there is no instrumental solution, can we develop an algorithm able to link spectral/spatial information between the different spectroscopic techniques?  How can we manage different spatial resolutions between imaging systems?  Are our actual chemometric methods adapted to big data? We will demonstrate in this thesis the potential of data fusion considering various spectroscopic techniques such as vibrational spectroscopy (Near infrared, Mid-infrared, Raman), Electron paramagnetic resonance (EPR), Laser-Induced Breakdown Spectroscopy (LIBS) and others from synchrotron beamlines.