Histone deacetylase inhibitors repress macrophage migration inhibitory factor (MIF) expression by targeting MIF gene transcription through a local chromatin deacetylation.

The cytokine macrophage migration inhibitory factor plays a central role in inflammation, cell proliferation and tumorigenesis. Moreover, macrophage migration inhibitory factor levels correlate with tumor aggressiveness and metastatic potential. Histone deacetylase inhibitors are potent antitumor agents recently introduced in the clinic. Therefore, we hypothesized that macrophage migration inhibitory factor would represent a target of histone deacetylase inhibitors. Confirming our hypothesis, we report that histone deacetylase inhibitors of various chemical classes strongly inhibited macrophage migration inhibitory factor expression in a broad range of cell lines, in primary cells and in vivo. Nuclear run on, transient transfection with macrophage migration inhibitory factor promoter reporter constructs and transduction with macrophage migration inhibitory factor expressing adenovirus demonstrated that trichostatin A (a prototypical histone deacetylase inhibitor) inhibited endogenous, but not episomal, MIF gene transcription. Interestingly, trichostatin A induced a local and specific deacetylation of macrophage migration inhibitory factor promoter-associated H3 and H4 histones which did not affect chromatin accessibility but was associated with an impaired recruitment of RNA polymerase II and Sp1 and CREB transcription factors required for basal MIF gene transcription. Altogether, this study describes a new molecular mechanism by which histone deacetylase inhibitors inhibit MIF gene expression, and suggests that macrophage migration inhibitory factor inhibition by histone deacetylase inhibitors may contribute to the antitumorigenic effects of histone deacetylase inhibitors.

Lessons learned from developing a national NCD programme in low- and middle-income countries (Seychelles)

This paper identifies selected issues and lessons learned from the implementation of a national program of prevention and control of non-communicable diseases (NCD) during the past 20 years in the Seychelles, a small island state in the African region. As early as in 1989, population-based surveys demonstrated high levels of several cardiovascular risk factors, which prompted an organized response by the government. The early creation of a NCD unit within the Ministry of Health, coupled with cooperation with international partners, enabled incremental capacity building and coherent development of NCD programs and policy. Information campaigns and screening for hypertension and diabetes in work/public places raised awareness and rallied increasingly broad awareness and support to NCD prevention and control. A variety of interventions were organized for tobacco control and comprehensive tobacco control legislation was enacted in 2009 (including total bans on tobacco advertising and on smoking in all enclosed public and work places). A recent School Nutrition Policy prohibits the sale of soft drinks in schools. At primary health care level, guidelines were developed for the management of hypertension and diabetes (these conditions are managed in all health centers within a national health system); regular interactive education sessions were organized for groups of high risk patients ("heart health club"); and specialized "NCD nurses" were trained. Decreasing prevalence of smoking is evidence of success, but the raising "diabesity epidemic" calls for strengthened health care to high-risk patients and broader multisectoral policy to mould an environment conducive to healthy behaviors. Key components of NCD prevention and control in Seychelles include effective surveillance mechanisms supplemented by focused research; generating broad interest and consensus on the need for prevention and control of NCD; mobilizing leadership and commitment at all levels; involving local and international expertise; building on existing efforts; and seeking integrated, multi-disciplinary and multisectoral approaches.

Combined transesophageal and surface MRI provides optimal imaging in aortic atherosclerosis.

OBJECTIVE: Surface magnetic resonance imaging (MRI) for aortic plaque assessment is limited by the trade-off between penetration depth and signal-to-noise ratio (SNR). For imaging the deep seated aorta, a combined surface and transesophageal MRI (TEMRI) technique was developed 1) to determine the individual contribution of TEMRI and surface coils to the combined signal, 2) to measure the signal improvement of a combined surface and TEMRI over surface MRI, and 3) to assess for reproducibility of plaque dimension analysis. METHODS AND RESULTS: In 24 patients six black blood proton-density/T2-weighted fast-spin echo images were obtained using three surface and one TEMRI coil for SNR measurements. Reproducibility of plaque dimensions (combined surface and TEMRI) was measured in 10 patients. TEMRI contributed 68% of the signal in the aortic arch and descending aorta, whereas the overall signal gain using the combined technique was up to 225%. Plaque volume measurements had an intraclass correlation coefficient of as high as 0.97. CONCLUSION: Plaque volume measurements for the quantification of aortic plaque size are highly reproducible for combined surface and TEMRI. The TEMRI coil contributes considerably to the aortic MR signal. The combined surface and TEMRI approach improves aortic signal significantly as compared to surface coils alone. CONDENSED ABSTRACT: Conventional MRI aortic plaque visualization is limited by the penetration depth of MRI surface coils and may lead to suboptimal image quality with insufficient reproducibility. By combining a transesophageal MRI (TEMRI) with surface MRI coils we enhanced local and overall image SNR for improved image quality and reproducibility.

Characterization of the effects of radiation therapy on the tumour microenvironment and their consequences on cancer progression

RESUME La radiothérapie est utilisée avec succès pour le traitement d'un grand nombre de pathologies tumorales (1). Cependant, les récidives post-actiniques sont associées à un risque accru de développer des métastases régionales et à distance (2, 3). La prise en charge de ce type de patients demeure insatisfaisante à l'heure actuelle, principalement parce que les mécanismes physio-pathologiques sous- sous-jacents restent mal compris. Etant donné le rôle primordial du stroma dans la progression tumorale (4) et l'importance des effets de la radiothérapie sur le micro-environnement des tumeurs (5), nous avons émis l'hypothèse que la radiothérapie pouvait engendrer des modifications stromales susceptibles de contribuer à l'émergence d'un phénotype tumoral plus agressif. Nous avons observé que l'exposition préalable d'un environnement tumoral à des radiations ionisantes engendre une inhibition locale et à long terme de l'angiogenèse. Cette inhibition conduit à la création d'un environnement tumoral hypoxique favorisant l'invasion et la métastatisation tumorale. Les mécanismes sous-jacents impliquent l'activation de gènes prométastatiques sous le contrôle du facteur de transcription HIF-1, ainsi que la sélection hypoxique de cellules hautement invasives et métastatiques. Par des analyses de profile d'expression génétique ainsi que par des analyses fonctionnelles, nous avons identifié la protéine matri-cellulaire CYR61 ainsi que ses partenaires d'interaction, les intégrines aVb5/aVb3, comme médiateurs importants de ces effets. De plus, une corrélation significative a également été trouvée entre le niveau d'expression de CYR61 et le taux d'hypoxie dans un grand nombre de carcinomes mammaires chez l'humain. Une association a aussi été observée entre le niveau d'expression de CYR61 et le pronostic de patientes souffrant d'un cancer du sein traité par chimiothérapie adjuvante. Globalement ces résultats identifient l'interaction entre la protéine CYR61 et ses récepteurs aVb5/aVb3 comme un mécanisme important du processus de métastatisation et en font une cible thérapeutique potentielle pour le traitement de patients souffrant d'une récidive tumorale après un traitement de radiothérapie. Finalement, bien que l'inhibition de l'angiogenèse soit locale dans ce cas particulier, nos résultats justifient une surveillance particulière des patients souffrant d'une pathologie tumorale et étant au bénéfice d'un traitement inhibiteur de l'angiogenèse. SUMMARY Radiotherapy is successfully used to treat a large variety of tumours (1 ). However, cancer patients experiencing local recurrent disease after radiation therapy are at increased risk of developing regional and distant metastasis (2, 3). The clinical management of this condition represents a difficult and challenging issue, mainly because the underlying physio-pathological mechanisms remain poorly understood. Given the well established role of the tumour stroma in promoting cancer progression (4) and since radiotherapy is known to persistently alter the tumour microenvironment (5), we hypothesized that ionising radiations may generate stromal modifications contributing to the metastatic spread of relapsing tumours. Here, we report that irradiation of the prospective tumour microenvironment promotes tumour invasion and metastasis through a mechanism of local and sustained impairment of angiogenesis leading to both HIF-1 dependent activation of pro-metastatic genes and hypoxia-mediated selection of highly metastatic tumour cell variants. Through gene expression profiling and functional experiments, we identified the matricellular signalling protein CYR61 and its interaction partners aVb5/ aVb3 integrins as critical mediators of these effects. Furthermore, we found a significant correlation between CYR61 expression and the hypoxic status of a large number of human mammary carcinomas. A positive correlation between increased levels of CYR61 expression and shorter relapse free survival was also identified in breast cancer patients treated with adjuvant chemotherapy. Together, these results identify CYR61 and aVb5/aVb3 integrins as critical mediators of metastasis and potential therapeutic targets to improve outcome in patients with post-radiation tumour recurrences. Finally, although inhibition of angiogenesis is local in this setting, our data warrant close monitoring of tumour progression in patients under anti-angiogenic therapy.

3D seismic facies characterization and geological patterns recognition (Australian North West Shelf)

EXECUTIVE SUMMARY This PhD research, funded by the Swiss Sciences Foundation, is principally devoted to enhance the recognition, the visualisation and the characterization of geobodies through innovative 3D seismic approaches. A series of case studies from the Australian North West Shelf ensures the development of reproducible integrated 3D workflows and gives new insight into local and regional stratigraphic as well as structural issues. This project was initiated in year 2000 at the Geology and Palaeontology Institute of the University of Lausanne (Switzerland). Several collaborations ensured the improvement of technical approaches as well as the assessment of geological models. - Investigations into the Timor Sea structural style were carried out at the Tectonics Special Research Centre of the University of Western Australia and in collaboration with Woodside Energy in Perth. - Seismic analysis and attributes classification approach were initiated with Schlumberger Oilfield Australia in Perth; assessments and enhancements of the integrated seismic approaches benefited from collaborations with scientists from Schlumberger Stavanger Research (Norway). Adapting and refining from "linear" exploration techniques, a conceptual "helical" 3D seismic approach has been developed. In order to investigate specific geological issues this approach, integrating seismic attributes and visualisation tools, has been refined and adjusted leading to the development of two specific workflows: - A stratigraphic workflow focused on the recognition of geobodies and the characterization of depositional systems. Additionally, it can support the modelling of the subsidence and incidentally the constraint of the hydrocarbon maturity of a given area. - A structural workflow used to quickly and accurately define major and secondary fault systems. The integration of the 3D structural interpretation results ensures the analysis of the fault networks kinematics which can affect hydrocarbon trapping mechanisms. The application of these integrated workflows brings new insight into two complex settings on the Australian North West Shelf and ensures the definition of astonishing stratigraphic and structural outcomes. The stratigraphic workflow ensures the 3D characterization of the Late Palaeozoic glacial depositional system on the Mermaid Nose (Dampier Subbasin, Northern Carnarvon Basin) that presents similarities with the glacial facies along the Neotethys margin up to Oman (chapter 3.1). A subsidence model reveals the Phanerozoic geodynamic evolution of this area (chapter 3.2) and emphasizes two distinct mode of regional extension for the Palaeozoic (Neotethys opening) and Mesozoic (abyssal plains opening). The structural workflow is used for the definition of the structural evolution of the Laminaria High area (Bonaparte Basin). Following a regional structural characterization of the Timor Sea (chapter 4.1), a thorough analysis of the Mesozoic fault architecture reveals a local rotation of the stress field and the development of reverse structures (flower structures) in extensional setting, that form potential hydrocarbon traps (chapter 4.2). The definition of the complex Neogene structural architecture associated with the fault kinematic analysis and a plate flexure model (chapter 4.3) suggest that the Miocene to Pleistocene reactivation phases recorded at the Laminaria High most probably result from the oblique normal reactivation of the underlying Mesozoic fault planes. This episode is associated with the deformation of the subducting Australian plate. Based on these results three papers were published in international journals and two additional publications will be submitted. Additionally this research led to several communications in international conferences. Although the different workflows presented in this research have been primarily developed and used for the analysis of specific stratigraphic and structural geobodies on the Australian North West Shelf, similar integrated 3D seismic approaches will have applications to hydrocarbon exploration and production phases; for instance increasing the recognition of potential source rocks, secondary migration pathways, additional traps or reservoir breaching mechanisms. The new elements brought by this research further highlight that 3D seismic data contains a tremendous amount of hidden geological information waiting to be revealed and that will undoubtedly bring new insight into depositional systems, structural evolution and geohistory of the areas reputed being explored and constrained and other yet to be constrained. The further development of 3D texture attributes highlighting specific features of the seismic signal, the integration of quantitative analysis for stratigraphic and structural processes, the automation of the interpretation workflow as well as the formal definition of "seismo-morphologic" characteristics of a wide range of geobodies from various environments would represent challenging examples of continuation of this present research. The 21st century will most probably represent a transition period between fossil and other alternative energies. The next generation of seismic interpreters prospecting for hydrocarbon will undoubtedly face new challenges mostly due to the shortage of obvious and easy targets. They will probably have to keep on integrating techniques and geological processes in order to further capitalise the seismic data for new potentials definition. Imagination and creativity will most certainly be among the most important quality required from such geoscientists.

Radiothérapie externe accélérée postopératoire des carcinomes épidermoïdes localement évolués de la sphère ORL: étude prospective de phase II [Prospective study of accelerated postoperative radiation therapy in patients with squamous-cell carcinoma of the head and neck]

PURPOSE: To assess the feasibility and efficacy of accelerated postoperative radiation therapy (RT) in patients with squamous-cell carcinoma of the head and neck (SCCHN). PATIENTS AND METHODS: Between December 1997 and July 2001, 68 patients (male to female ratio: 52/16; median age: 60-years (range: 43-81) with pT1-pT4 and/or pN0-pN3 SCCHN (24 oropharynx, 19 oral cavity, 13 hypopharynx, 5 larynx, 3 unknown primary, 2 maxillary sinus, and 2 salivary gland) were included in this prospective study. Postoperative RT was indicated because extracapsular infiltration (ECI) was observed in 20 (29%), positive surgical margins (PSM) in 20 (29%) or both in 23 patients (34%). Treatment consisted of external beam RT 66 Gy in 5 weeks and 3 days. Median follow-up was 15 months. RESULTS: According to CTC 2.0, acute morbidity was acceptable: grade 3 mucositis was observed in 15 (22%) patients, grade 3 dysphagia in 19 (28%) patients, grade 3 skin erythema in 21 (31%) patients with a median weight loss of 3.1 kg (range: 0-16). No grade 4 toxicity was observed. Median time to relapse was 13 months; we observed only three (4%) local and four (6%) regional relapses, whereas eight (12%) patients developed distant metastases without any evidence of locoregional recurrence. The 2 years overall-, disease-free survival, and actuarial locoregional control rates were 85, 73 and 83% respectively. CONCLUSION: The reduction of the overall treatment time using postoperative accelerated RT with weekly concomitant boost (six fractions per week) is feasible with local control rates comparable to that of published data. Acute RT-related morbidity is acceptable.

Waveform inversion of crosshole georadar data: influence of source wavelet variability and the suitability of a single wavelet assumption

Waveform-based tomographic imaging of crosshole georadar data is a powerful method to investigate the shallow subsurface because of its ability to provide images of electrical properties in near-surface environments with unprecedented spatial resolution. A critical issue with waveform inversion is the a priori unknown source signal. Indeed, the estimation of the source pulse is notoriously difficult but essential for the effective application of this method. Here, we explore the viability and robustness of a recently proposed deconvolution-based procedure to estimate the source pulse during waveform inversion of crosshole georadar data, where changes in wavelet shape with location as a result of varying near-field conditions and differences in antenna coupling may be significant. Specifically, we examine whether a single, average estimated source current function can adequately represent the pulses radiated at all transmitter locations during a crosshole georadar survey, or whether a separate source wavelet estimation should be performed for each transmitter gather. Tests with synthetic and field data indicate that remarkably good tomographic reconstructions can be obtained using a single estimated source pulse when moderate to strong variability exists in the true source signal with antenna location. Only in the case of very strong variability in the true source pulse are tomographic reconstructions clearly improved by estimating a different source wavelet for each transmitter location.

Active learning methods for remote sensing image classification

In this paper, we propose two active learning algorithms for semiautomatic definition of training samples in remote sensing image classification. Based on predefined heuristics, the classifier ranks the unlabeled pixels and automatically chooses those that are considered the most valuable for its improvement. Once the pixels have been selected, the analyst labels them manually and the process is iterated. Starting with a small and nonoptimal training set, the model itself builds the optimal set of samples which minimizes the classification error. We have applied the proposed algorithms to a variety of remote sensing data, including very high resolution and hyperspectral images, using support vector machines. Experimental results confirm the consistency of the methods. The required number of training samples can be reduced to 10% using the methods proposed, reaching the same level of accuracy as larger data sets. A comparison with a state-of-the-art active learning method, margin sampling, is provided, highlighting advantages of the methods proposed. The effect of spatial resolution and separability of the classes on the quality of the selection of pixels is also discussed.

Fuel metabolism during exercise in euglycaemia and hyperglycaemia in patients with type 1 diabetes mellitus--a prospective single-blinded randomised crossover trial.

Aims/hypothesis We assessed systemic and local muscle fuel metabolism during aerobic exercise in patients with type I diabetes at euglycaemia and hyperglycaemia with identical insulin levels.Methods This was a single-blinded randomised crossover study at a university diabetes unit in Switzerland. We studied seven physically active men with type I diabetes (mean +/- SEM age 33.5 +/- 2.4 years, diabetes duration 20.1 +/- 3.6 years, HbA(1c) 6.7 +/- 0.2% and peak oxygen uptake [VO2peak] 50.3 +/- 4.5 ml min(-1) kg(-1)). Men were studied twice while cycling for 120 min at 55 to 60% of VO2peak, with a blood glucose level randomly set either at 5 or 11 mmol/l and identical insulinaemia. The participants were blinded to the glycaemic level; allocation concealment was by opaque, sealed envelopes. Magnetic resonance spectroscopy was used to quantify intramyocellular glycogen and lipids before and after exercise. Indirect calorimetry and measurement of stable isotopes and counter-regulatory hormones complemented the assessment of local and systemic fuel metabolism.Results The contribution of lipid oxidation to overall energy metabolism was higher in euglycaemia than in hyperglycaemia (49.4 +/- 4.8 vs 30.6 +/- 4.2%; p<0.05). Carbohydrate oxidation accounted for 48.2 +/- 4.7 and 66.6 +/- 4.2% of total energy expenditure in euglycaemia and hyperglycaemia, respectively (p<0.05). The level of intramyocellular glycogen before exercise was higher in hyperglycaemia than in euglycaemia (3.4 +/- 0.3 vs 2.7 +/- 0.2 arbitrary units [AU]; p<0.05). Absolute glycogen consumption tended to be higher in hyperglycaemia than in euglycaemia (1.3 +/- 0.3 vs 0.9 +/- 0.1 AU). Cortisol and growth hormone increased more strongly in euglycaemia than in hyperglycaemia (levels at the end of exercise 634 52 vs 501 +/- 32 nmol/l and 15.5 +/- 4.5 vs 7.4 +/- 2.0 ng/ml, respectively; p<0.05).Conclusions/interpretation Substrate oxidation in type I diabetic patients performing aerobic exercise in euglycaemia is similar to that in healthy individuals revealing a shift towards lipid oxidation during exercise. In hyperglycaemia fuel metabolism in these patients is dominated by carbohydrate oxidation. Intramyocellular glycogen was not spared in hyperglycaemia.

Coexistence of specialist and generalist species is shaped by dispersal and environmental factors.

Disentangling the mechanisms mediating the coexistence of habitat specialists and generalists has been a long-standing subject of investigation. However, the roles of species traits and environmental and spatial factors have not been assessed in a unifying theoretical framework. Theory suggests that specialist species are more competitive in natural communities. However, empirical work has shown that specialist species are declining worldwide due to habitat loss and fragmentation. We addressed the question of the coexistence of specialist and generalist species with a spatially explicit metacommunity model in continuous and heterogeneous environments. We characterized how species' dispersal abilities, the number of interacting species, environmental spatial autocorrelation, and disturbance impact community composition. Our results demonstrated that species' dispersal ability and the number of interacting species had a drastic influence on the composition of metacommunities. More specialized species coexisted when species had large dispersal abilities and when the number of interacting species was high. Disturbance selected against highly specialized species, whereas environmental spatial autocorrelation had a marginal impact. Interestingly, species richness and niche breadth were mainly positively correlated at the community scale but were negatively correlated at the metacommunity scale. Numerous diversely specialized species can thus coexist, but both species' intrinsic traits and environmental factors interact to shape the specialization signatures of communities at both the local and global scales.

Semi-supervised remote sensing image classification with cluster kernels

A semisupervised support vector machine is presented for the classification of remote sensing images. The method exploits the wealth of unlabeled samples for regularizing the training kernel representation locally by means of cluster kernels. The method learns a suitable kernel directly from the image and thus avoids assuming a priori signal relations by using a predefined kernel structure. Good results are obtained in image classification examples when few labeled samples are available. The method scales almost linearly with the number of unlabeled samples and provides out-of-sample predictions.

Kernel-based mapping of orographic rainfall enhancement in the Swiss Alps as detected by weatherradar

In this paper, we develop a data-driven methodology to characterize the likelihood of orographic precipitation enhancement using sequences of weather radar images and a digital elevation model (DEM). Geographical locations with topographic characteristics favorable to enforce repeatable and persistent orographic precipitation such as stationary cells, upslope rainfall enhancement, and repeated convective initiation are detected by analyzing the spatial distribution of a set of precipitation cells extracted from radar imagery. Topographic features such as terrain convexity and gradients computed from the DEM at multiple spatial scales as well as velocity fields estimated from sequences of weather radar images are used as explanatory factors to describe the occurrence of localized precipitation enhancement. The latter is represented as a binary process by defining a threshold on the number of cell occurrences at particular locations. Both two-class and one-class support vector machine classifiers are tested to separate the presumed orographic cells from the nonorographic ones in the space of contributing topographic and flow features. Site-based validation is carried out to estimate realistic generalization skills of the obtained spatial prediction models. Due to the high class separability, the decision function of the classifiers can be interpreted as a likelihood or susceptibility of orographic precipitation enhancement. The developed approach can serve as a basis for refining radar-based quantitative precipitation estimates and short-term forecasts or for generating stochastic precipitation ensembles conditioned on the local topography.

Key electrophysiological, molecular, and metabolic signatures of sleep and wakefulness revealed in primary cortical cultures.

Although sleep is defined as a behavioral state, at the cortical level sleep has local and use-dependent features suggesting that it is a property of neuronal assemblies requiring sleep in function of the activation experienced during prior wakefulness. Here we show that mature cortical cultured neurons display a default state characterized by synchronized burst-pause firing activity reminiscent of sleep. This default sleep-like state can be changed to transient tonic firing reminiscent of wakefulness when cultures are stimulated with a mixture of waking neurotransmitters and spontaneously returns to sleep-like state. In addition to electrophysiological similarities, the transcriptome of stimulated cultures strikingly resembles the cortical transcriptome of sleep-deprived mice, and plastic changes as reflected by AMPA receptors phosphorylation are also similar. We used our in vitro model and sleep-deprived animals to map the metabolic pathways activated by waking. Only a few metabolic pathways were identified, including glycolysis, aminoacid, and lipids. Unexpectedly large increases in lysolipids were found both in vivo after sleep deprivation and in vitro after stimulation, strongly suggesting that sleep might play a major role in reestablishing the neuronal membrane homeostasis. With our in vitro model, the cellular and molecular consequences of sleep and wakefulness can now be investigated in a dish.

Advanced Kernel Methods For Remote Sensing Image Classification

Résumé Suite aux recentes avancées technologiques, les archives d'images digitales ont connu une croissance qualitative et quantitative sans précédent. Malgré les énormes possibilités qu'elles offrent, ces avancées posent de nouvelles questions quant au traitement des masses de données saisies. Cette question est à la base de cette Thèse: les problèmes de traitement d'information digitale à très haute résolution spatiale et/ou spectrale y sont considérés en recourant à des approches d'apprentissage statistique, les méthodes à noyau. Cette Thèse étudie des problèmes de classification d'images, c'est à dire de catégorisation de pixels en un nombre réduit de classes refletant les propriétés spectrales et contextuelles des objets qu'elles représentent. L'accent est mis sur l'efficience des algorithmes, ainsi que sur leur simplicité, de manière à augmenter leur potentiel d'implementation pour les utilisateurs. De plus, le défi de cette Thèse est de rester proche des problèmes concrets des utilisateurs d'images satellite sans pour autant perdre de vue l'intéret des méthodes proposées pour le milieu du machine learning dont elles sont issues. En ce sens, ce travail joue la carte de la transdisciplinarité en maintenant un lien fort entre les deux sciences dans tous les développements proposés. Quatre modèles sont proposés: le premier répond au problème de la haute dimensionalité et de la redondance des données par un modèle optimisant les performances en classification en s'adaptant aux particularités de l'image. Ceci est rendu possible par un système de ranking des variables (les bandes) qui est optimisé en même temps que le modèle de base: ce faisant, seules les variables importantes pour résoudre le problème sont utilisées par le classifieur. Le manque d'information étiquétée et l'incertitude quant à sa pertinence pour le problème sont à la source des deux modèles suivants, basés respectivement sur l'apprentissage actif et les méthodes semi-supervisées: le premier permet d'améliorer la qualité d'un ensemble d'entraînement par interaction directe entre l'utilisateur et la machine, alors que le deuxième utilise les pixels non étiquetés pour améliorer la description des données disponibles et la robustesse du modèle. Enfin, le dernier modèle proposé considère la question plus théorique de la structure entre les outputs: l'intègration de cette source d'information, jusqu'à présent jamais considérée en télédétection, ouvre des nouveaux défis de recherche. Advanced kernel methods for remote sensing image classification Devis Tuia Institut de Géomatique et d'Analyse du Risque September 2009 Abstract The technical developments in recent years have brought the quantity and quality of digital information to an unprecedented level, as enormous archives of satellite images are available to the users. However, even if these advances open more and more possibilities in the use of digital imagery, they also rise several problems of storage and treatment. The latter is considered in this Thesis: the processing of very high spatial and spectral resolution images is treated with approaches based on data-driven algorithms relying on kernel methods. In particular, the problem of image classification, i.e. the categorization of the image's pixels into a reduced number of classes reflecting spectral and contextual properties, is studied through the different models presented. The accent is put on algorithmic efficiency and the simplicity of the approaches proposed, to avoid too complex models that would not be used by users. The major challenge of the Thesis is to remain close to concrete remote sensing problems, without losing the methodological interest from the machine learning viewpoint: in this sense, this work aims at building a bridge between the machine learning and remote sensing communities and all the models proposed have been developed keeping in mind the need for such a synergy. Four models are proposed: first, an adaptive model learning the relevant image features has been proposed to solve the problem of high dimensionality and collinearity of the image features. This model provides automatically an accurate classifier and a ranking of the relevance of the single features. The scarcity and unreliability of labeled. information were the common root of the second and third models proposed: when confronted to such problems, the user can either construct the labeled set iteratively by direct interaction with the machine or use the unlabeled data to increase robustness and quality of the description of data. Both solutions have been explored resulting into two methodological contributions, based respectively on active learning and semisupervised learning. Finally, the more theoretical issue of structured outputs has been considered in the last model, which, by integrating outputs similarity into a model, opens new challenges and opportunities for remote sensing image processing.