Assous, Saïd (2005) Time-frequency analysis by the S transform and interpretation of the laser Doppler flowmetry signals: clinic diagnosis applications. PhD thesis Automatique et traitement du signal, ENSAM 2005ENAM0026.
Abstract
aThe laser Doppler flowmetry (LDF) technique is based on the spectral broadening of monochromatic light, that interacts with moving red blood cells in tissue. The power spectral density of the backscattered light can be processed to yield an estimate of microvascular tissue perfusion in the form of a signal. The primary objective of this work is to explore the information contained in this signal in order to understand the physiological phenomena which appear in the microcirculation and to present solutions of assistance in the diagnosis of corresponding pathologies. The decision problem may be divided into the stages of feature extraction, dimensionality reduction and pattern recognition. This work shows that decision performance depends largely upon the signal representation. The study relies on the S transform, a linear time-frequency representation which overcomes the short time Fourier transform fixed length window limitation and the phase notion in the wavelet transform for non stationary signal analysis. This transform provides a powerful framework for feature extraction, localizing the discriminant information in the laser Doppler flowmetry signal in time and in frequency. This new approach makes possible the analysis of the five characteristic frequencies contained in laser Doppler flowmetry signals and contributes to the apprehension of the signals of reactive hyperaemia. In addition, due to the high dimension of time-frequency representations, its success relies upon an appropriate form of dimensionality reduction. It is shown that the singular values decomposition provides an effective means of concentrating that information which is important, and discarding that which is irrelevant. Its singular vectors reflect the behavior spectral in the time of the various physiological activities present in the laser Doppler flowmetry signal.
| Item Type: | PhD Thesis (PhD) |
|---|
| Thesis Supervisor: | L'Huillier, Jean-Pierre |
|---|
| Date: | December 2005 |
|---|
| Board of examiners: | Chapeau-Blondeau, François and Harba, Rachid and Humeau, Anne and L'Huillier, Jean-Pierre and Martin, Nadine and Wolf, Didier |
|---|
| Ecole Doctorale: | ED 432 ECOLE DOCTORALE SCIENCES DES METIERS DE L'INGENIEUR |
|---|
| Discipline: | Automatique et traitement du signal |
|---|
| Collection (Fonds): | ENSAM |
|---|
| Institution: | ENSAM |
|---|
| Subjects: | 2. Information and Communication Sciences and Technologies |
|---|
| Uncontrolled Keywords: | Laser Doppler flowmetry, Time-frequency analysis, S transform, Feature extraction, Singular values decomposition, Diagnosis, Fluxmétrie laser Doppler, Analyse temps-fréquence, transformée en S, Extraction de caractéristiques, Décomposition en valeurs singulières, Diagnostic |
|---|
Statistiques de consultation
Repository Staff Only: edit this item
