Cette thèse porte sur l'étude de la méthode d'imagerie photoacoustique, une nouvelle modalité hybride permettant de combiner la haute résolution de l'imagerie par ultrasons et le contraste de l'imagerie optique. Nous y étudions en particulier le problème inverse associé et sa résolution : il se décompose en l'inversion de l'équation d'ondes et en celle de l'équation de diffusion optique, dont le but est de retrouver les paramètres optiques du milieu. Dans la première partie de cette étude nous développons un modèle permettant de prendre en compte les variations de la vitesse acoustique dans le milieu biologique. En effet, la plupart des méthodes d'inversion supposent une vitesse acoustique constante, ce qui est à l'origine d'erreurs dans les reconstructions. La deuxième partie de la thèse porte sur une étude mathématique du phénomène de limitation de la profondeur de l'imagerie photoacoustique. Nous calculons une estimation de stabilité du problème inverse dans le cas d'un milieu stratifié et nous montrons que la reconstruction se dégrade avec la profondeur. Nous étudions dans la dernière partie le phénomène photoacoustique en présence de nanoparticules métalliques : ces marqueurs permettent d'amplifier par des résonances le signal photoacoustique généré autour d'elles. Elles permettent ainsi une meilleure visibilité des tissus en profondeur. Nous explicitons ici le modèle mathématique de génération du signal photoacoustique, ainsi que la résolution théorique du problème inverse photoacoustique dans ce contexte. / This thesis work is related to photoacoustic imaging techniques which are new multiwave modalities in medical imaging that combine both high resolution of ultrasounds and contrast of optical methods. Weprecisely studied the inverse problem that consists of determining the optical coefficients of biologicaltissues from measurement of acoustic waves generated by the photoacoustic effect. The photoacoustic inverse problem proceeds in two steps.We first retrieve the initial pressure from the measurement of the pressure wave on a part of the boundary of the sample. The first inversion takes then the form of a linear inverse source problem and provides internal data for the optical waves that are more sensitive to the contrast of the absorption and diffusion coefficients. In a second step we recover the optical coefficients from the acquired internal data.The aim of this work is to study the two inversions in different contexts. In the first part, we develop a model that takes into account the variations of the acoustic speed in the medium. Indeed, most of the inversion methods suppose that the acoustic speed is constant, and this assumption can lead to errors in the reconstruction of the optical coefficients. The second part of this work is the derivation of stability estimates for the photoacoustic inverse problem in a layered medium. We prove that the reconstruction is getting worse with depth. This is one of the main drawbacks of the photacoustic method, the imaging depth is limited to a few centimeters. The last part is about photoacoustic generation with plasmonic nanoparticles. They enhance the photoacoustic signal around them, so that we can investigate the tissue more deeply. We derive the mathematical model of the photoacoustic generation by heating nanoparticles, and we solve the photoacoustic inverse problem in this context.

Abstract The thesis deals with numerical methods designed for the modeling and inversion of geophysical electromagnetic (EM) measurements using a conductive thin plate model. The main objectives are to study the EM induction problem in general and to develop practical interpretation tools for mineral prospecting in particular. The starting point is a linearized inversion method based on the singular value decomposition and a new adaptive damping method. The inversion method is introduced to the interpretation of time-domain EM (TEM) measurements using a thin plate in free-space. The central part of the thesis is a new approximate modeling method, which is based on an integral equation approach and a special lattice model. At first the modeling method is applied to the interpretation of frequency-domain EM (FEM) data using a thin plate in conductive two-layered earth. After this time-domain responses are modeled applying a Fourier-sine transform of broadband FEM computations. The results demonstrate that the approximate computational method can model the geophysical frequency and time-domain EM responses of a thin conductor in conductive host medium with sufficient accuracy, and that the inversion method can provide reliable estimates for the model parameters. The fast forward computation enables interactive interpretation of FEM data and feasible forward modeling of TEM responses. The misfit function mapping and analysis of the singular value decomposition have provided additional information about the sensitivity, resolution, and the correlation behavior of the thin plate parameters.

Electrical impedance tomography of soft tissue: Forward and inverse modelling The diploma thesis builds the neccesary apparatus to formulate and solve the inverse problem of Eletrical Impedance Tomography (EIT), including strategies to remedy the ill-conditioning of the problem. The problem itself lies in determining the structure of a body of interest by driving a set of electrical currents through electrodes connected to its surface. The aim of the thesis is to investigate possible utility of this method in medical applications, namely scanning for malignancies in the female breast, by studying the interaction of tissue with the electromagnetic field and by preparing a set of correspoding numerical experiments. An approximate characterization of the method's sensitivity with respect to noise is derived based on the most basic set of such numerical experiments, which were prepared by a complete software solution called prs4D developed by the author and his advisor, while some aspects of its implementation are included in the thesis.

In this thesis we generalize the problem of phase retrieval of vector to that of multi-vector. The identification of the multi-vector is done up to some special classes of isometries in the space. We give some upper and lower estimates on the minimal number of multi-linear operators needed for the retrieval. The results are preliminary and far from sharp.

Les problèmes inverses de localisation de sources en électroencéphalographie (EEG) consistent à retrouver le lieu d'origine dans le cerveau des signaux mesurés sur le scalp. La qualité du résultat de localisation dépend des modèles géométriques et de conductivité électrique utilisés pour la résolution du problème. Parmi les tissus composant la tête, le crâne est celui dont la conductivité est la plus influente, en particulier à cause de sa faible valeur. De plus, le crâne humain est un tissu osseux comportant des parties dures et spongieuses, d'épaisseurs variables. Sa composition est très variable selon les individus, en termes de géométrie et de valeurs des conductivités, d'où la nécessité de développer des technique d'estimation de conductivités inconnues dans le crâne. Le but de cette thèse est de réduire l'incertitude sur la conductivité du crâne, pour des géométries sphériques et réalistes, en particulier en vue d’améliorer les résultats d'estimation des sources dans le problème inverse EEG. Dans le cas d'un domaine sphérique à 3 couches, l'existence, l'unicité et la stabilité de la conductivité dans la couche intermédiaire (crâne) sont discutées, et une procédure de reconstruction est proposée. Puis deux modèles plus réalistes de tête sont étudiés, l'un pour lequel le crâne est modelisé par un seul compartiment, l'autre dans lequel les parties spongieuses et dure sont distinguées. Des simulations numériques mettent en évidence le rôle de la structure interne du crâne pour la détermination de sa conductivité. / One of the major issues related to electroencephalography (EEG) is to localize where in the brain signals are generated, this is so called inverse problem of source localization. The quality of the source localization depends on the accuracy of the geometry and the electrical conductivity model used to solve the problem. Among the head tissues, the skull conductivity is the one that influences most the accuracy of the source localization, due to its low value. Moreover, the human skull is a bony tissue consisting of compact and spongy bone layers, whose thickness vary across the skull. As the skull tissue composition has strong inter-individual variability both in terms of geometry and of individual conductivity, conductivity estimation techniques are required in order to determine the unknown skull conductivity. The aim of this thesis is to reduce the uncertainty on the skull conductivity both in spherical and realistic head geometries in order to increase the quality of the inverse source localization problem. Therefore, conductivity estimation is first performed on a 3-layered spherical head model. Existence, uniqueness and stability of the conductivity in the intermediate skull layer are discussed, together with a constructive recovery scheme. Then a simulation study is performed comparing two realistic head models, a bulk model where the skull is modelled as a single compartment and a detailed one accounting for the compact and spongy bone layers, in order to determine the importance of the internal skull structure for conductivity estimation in EEG.

Ablative materials can sustain very high temperatures in which surface thermochemical processes are significant enough to cause surface recession. Existence of moving boundary over a wide range of temperatures, temperature-dependent thermophysical properties of ablators, and no prior knowledge about the location of the moving surface augment the difficulty for predicting the exposed heat flux at the receding surface of ablators. In this paper, the conjugate gradient method is proposed to estimate the unknown surface recession and time-varying net surface heat flux for these kinds of problems. The first order Tikhonov regularization is employed to stabilize the inverse solution. Considering the complicated phenomena that are taking place, it is shown via simulated experiment that unknown quantities can be obtained with reasonable accuracy using this method despite existing noises in the measurement data.

In this thesis, we present a sensor model comprised of a geometrically nonlinear beam coupled with a nonlinear viscoelastic Pasternak foundation via a distributed system of compliant elements. The governing equations of the system are obtained. By posing an inverse problem, the model is used to simulate the estimation of coupled substrates' material (constitutive) parameters. In the inverse problem, beam deformations are considered as measured parameters, and therefore an eventual hardware implementation would require measurements of these quantities. Different case studies are simulated to assess the robustness and applicability of this sensor model.

No description available.

In this thesis, the perturbation-based decomposition technique developed by Szlavik [1] was used in an attempt to solve the inverse problem in EEG source localization. A set of dipole locations were forward modeled using a 4-layer sphere model of the head at uniformly distributed lead locations to form the vector basis necessary for the method. Both a two-dimensional and a pseudo-three-dimensional versions of the model were assessed with the two-dimensional model yielding decompositions with minimal error and the pseudo-three-dimensional version having unacceptable levels of error. The utility of interpolation as a method to reduce the number of data points to become overdefined was assessed as well. The approach was effective as long as the number of component functions did not exceed the number of data points and stayed relatively small (less than 77 component functions). This application of the method to a spatially variate system indicates its potential for other systems and with some tweaking to the least squares algorithm used, could be applied to multivariate systems.

No description available.