Quantification en imagerie optique diffuse cerebrale: Analyse du signal et etude du probleme direct.

摘要

The aim of this PhD thesis lies in the development of multimodal data fusion methods to improve signal analysis and image quantification in diffuse optical imaging. The multimodal fusion involves two data types: functional data from diffuse optical imaging (DOI) and anatomical data from magnetic resonance imaging (MRI). In this context, high MRI spatial resolution is combined with high DOI temporal resolution to allow the temporal analysis of DOI signals and their localization within the brain.;The third part presents to the diffuse optical theory and the development of DOI signal analysis methods. We hypothesize that signal analysis can enable physiological signal source differentiations and improve cerebral activity detection. Since injected light crosses several head tissues before being detected, many informations (time and frequency) are brought on extra- and intracerebral physiology. Given this spectral structure, it can be advantageous to describe the signal in the time-frequency space by wavelets extensions . In this part, it is showed that physiology emerges naturally in the time-frequency plane and can be distinguished more readily than by a standard Fourier analysis. The ability of analytical wavelets to define an instantaneous phase in a concrete manner opens the door for a new measure: the phase-lock (synchrony) of the signal with itself. Moreover, analysis techniques developed in this work have allowed the characterization physiological noise and the estimation of the strength of hemodynamic response in diffuse optical imaging. Analysis performed on experimental DOI data provided us with a quantitative measure of the 1/f noise for hemoglobin concentration measures.;The multimodal data fusion of DOI functional and MRI anatomical data in the fourth part. Here, algorithms are developed to localize optical probes outside the MRI scanner. The integration of neuronavigation tools and a priori anatomical MRI data can improve the optical probes positioning. It is demonstrated that displaying optical sensitivity on MRI images can improve significantly the optical configuration positioning and hence the resulting imagery.;The fifth part exposes the new hybrid method for describing the photons propagation in a multilayered medium with the boundary element method (BEM). The method is formulated using the Born approximation applied to the diffusion equation which is defined by an absorption perturbation. This is relevant with cerebral DOI since detected absorption changes reflect hemoglobin changes involving in the hemodynamic response function. The integral formulation of the forward model with this method enables the image quantification. This model necessitates elaboration of segmentation tools of cerebral tissues from a priori MRI anatomical data. Indeed, these MRI segmentations enable to define distinct optical properties for each tissue. The algorithms are written to follow the hybrid approach: they can accomplish boundary segmentations for the need of the BEM and also volumetric segmentations necessary for the perturbation definition. Results are validated with Monte Carlo simulations. These methods describe the radiative transport equation and are computationally intensive. Results suggest that 3D multilayered medium diffuse optical tomography can be performed using the perturbative BEM approach with a reduced computational cost. However, this method cannot be applied in medium with low- or non-scattering regions and close (∼ 2 mm) to the source since the diffusion equation is not valid. This is the case in diffuse optical imaging when considering the cerebral spinal fluid (CSF) as a sub-region. In addition, in the context of 3D cerebral imaging, BEM enables to build accurate surface tissue meshes. This is not true with volumetric meshes built with FEM since modeling precisely the cerebral convolutions on the surface of the brain could be particulary arduous. Finally, the new hybrid method is less computationally expensive than Monte Carlo simulations.;The first part of the thesis introduces the work by presenting hypothesis, objectives, methods, conclusions and original scientific contributions. The thesis orientation is also included to help the reader to navigate through the thesis. Second part exposes complex biological mechanisms involving in the hemodynarnical response function.;Some research directions are left unanswered. Briefly, it could be interesting to investigate on new spectral methods to analyse physiological signals. Indeed, the understanding of physiology behaviors undoubtedly leads to improvements of hemoglobin concentrations estimation. In the case of the forward model definition, the addition of diffusion MRI could help in modeling the anisotropic photon propagation within the fibers of the white matter. (Abstract shortened by UMI.)