Localization of embedded inclusions using detection of fluorescence: Feasibility study based on simulation data, LS-SVM modeling and EPO pre-processing

Fablen Chauchard, Jenny Svensson, Johan Axelsson, Stefan Andersson-Engels, Sylvie Roussel

Research output: Contribution to journalArticlepeer-review

Abstract

Fluorescence spectroscopy is a useful technique for tissue diagnostics and is also a promising tool in the characterization of embedded structures in tissue. The emitted fluorescence from an embedded inclusion, marked with a fluorescent compound, is affected by several factors as the light propagates through the medium to the tissue boundary, where the fluorescence light is detected. Tissue absorption, scattering and autofluorescence, as well as the size and depth of the inclusion, affect the detected fluorescence light. The aim of this study is to investigate if the size and location of a fluorescent inclusion could be determined using models based a combination of External Parameter Orthogonalisation (EPO) and Least Squares Support Vector Machine (LS-SVM). This can be very useful for data pre-processing before a full fluorescence tomography reconstruction. The data set consisted of simulated multispectral fluorescence, where depth and radius of a spherical fluorescent inclusion were varied as well as the fluorescence contrast and optical properties of the surrounding tissue. The results showed that the non-linear models based on LS-SVM can simultaneously predict both radius and depth. It was observed that EPO acts as a useful pre-processing tool on spectra for this nonlinear model and that it was necessary to perform EPO to be able to predict the depth with the LS-SVM model. (C) 2007 Elsevier B.V. All rights reserved.
Original languageEnglish
Pages (from-to)34-42
JournalChemometrics and Intelligent Laboratory Systems
Volume91
Issue number1
DOIs
Publication statusPublished - 2008

Subject classification (UKÄ)

  • Atom and Molecular Physics and Optics

Free keywords

  • embedded
  • non-linearity
  • LS-SVM
  • external parameter orthogonalisation
  • multivariate analysis
  • multispectral
  • fluorescence spectroscopy
  • lesions
  • fluorescence tomography

Fingerprint

Dive into the research topics of 'Localization of embedded inclusions using detection of fluorescence: Feasibility study based on simulation data, LS-SVM modeling and EPO pre-processing'. Together they form a unique fingerprint.

Cite this