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.

Estimating and explaining the effect of education and income on head and neck cancer risk : INHANCE consortium pooled analysis of 31 case-control studies from 27 countries

Low socioeconomic status has been reported to be associated with head and neck cancer risk. However, previous studies have been too small to examine the associations by cancer subsite, age, sex, global region, and calendar time, and to explain the association in terms of behavioural risk factors. Individual participant data of 23,964 cases with head and neck cancer and 31,954 controls from 31 studies in 27 countries pooled with random effects models. Overall, low education was associated with an increased risk of head and neck cancer (OR = 2·50; 95%CI 2·02- 3·09). Overall one-third of the increased risk was not explained by differences in the distribution of cigarette smoking and alcohol behaviours; and it remained elevated among never users of tobacco and non-drinkers (OR = 1·61; 95%CI 1·13 - 2·31). More of the estimated education effect was not explained by cigarette smoking and alcohol behaviours: in women than in men, in older than younger groups, in the oropharynx than in other sites, in South/Central America than in Europe/North America, and was strongest in countries with greater income inequality. Similar findings were observed for the estimated effect of low vs high household income. The lowest levels of income and educational attainment were associated with more than 2-fold increased risk of head and neck cancer, which is not entirely explained by differences in the distributions of behavioural risk factors for these cancers, and which varies across cancer sites, sexes, countries, and country income inequality levels. © 2014 Wiley Periodicals, Inc.

PTLD Burkitt Lymphoma in a Patient with Remote Lymphomatoid Granulomatosis.

Posttransplant lymphoproliferative disorder (PTLD) is a potentially fatal complication of solid organ transplantation. The majority of PTLD is of B-cell origin, and 90% are associated with the Epstein-Barr virus (EBV). Lymphomatoid granulomatosis (LG) is a rare, EBV-associated systemic angiodestructive lymphoproliferative disorder, which has rarely been described in patients with renal transplantation. We report the case of a patient with renal transplantation for SLE, who presented, 9 months after renal transplantation, an EBV-associated LG limited to the intracranial structures that recovered completely after adjustment of her immunosuppressive treatment. Nine years later, she developed a second PTLD disorder with central nervous system initial manifestation. Workup revealed an EBV-positive PTLD Burkitt lymphoma, widely disseminated in most organs. In summary, the reported patient presented two lymphoproliferative disorders (LG and Burkitt's lymphoma), both with initial neurological manifestation, at 9 years interval. With careful reduction of the immunosuppression after the first manifestation and with the use of chemotherapy combined with radiotherapy after the second manifestation, our patient showed complete disappearance of neurologic symptoms and she is clinically well with good kidney function. No recurrence has been observed by radiological imaging until now.

Oligoanalgesia in the emergency department: short-term beneficial effects of an education program on acute pain.

STUDY OBJECTIVE: Acute pain is the most frequent complaint in emergency department (ED) admissions, but its management is often neglected, placing patients at risk of oligoanalgesia. We evaluate the effect of the implementation of guidelines for pain management in ED patients with pain at admission or anytime during their stay in our ED. METHODS: This prospective pre-post intervention cohort study included data collection both before and after guideline implementation. Consecutive adult patients admitted with acute pain from any cause or with pain at any time after admission were enrolled. The quality of pain management was evaluated according to information in the ED medical records by using a standardized collection form, and its impact on patients was recorded with a questionnaire at discharge. RESULTS: Two hundred forty-nine and 192 patients were included during pre- and postintervention periods. Pain was documented in 61% and 76% of nurse and physician notes, respectively, versus 78% and 85% after the intervention (difference 17%/9%; 95% confidence interval [CI] 8% to 26%/2% to 17%, respectively). Administration of analgesia increased from 40% to 63% (difference 23%; 95% CI 13% to 32%) and of morphine from 10% to 27% (difference 17%; 95% CI 10% to 24%). Mean doses of intravenous morphine increased from 2.4 mg (95% CI 1.9 to 2.9 mg) to 4.6 mg (95% CI 3.9 to 5.3 mg); administration of nonsteroidal antiinflammatory drugs and acetaminophen increased as well. There was a greater reduction of visual analogue scale score after intervention: 2.1 cm (95% CI 1.7 to 2.4 cm) versus 2.9 cm (95% CI 2.5 to 3.3 cm), which was associated with improved patient satisfaction. CONCLUSION: Education program and guidelines implementation for pain management lead to improved pain management, analgesia, and patient satisfaction in the ED.

E-tools for civic education : experiences from the election campaign in Switzerland 2007

Political participation is often very low in Switzerland especially among students and young citizens. In the run-up to the Swiss parliamentary election in October 2007 several online tools and campaigns were developed with the aim to increase not only the level of information about the political programs of parties and candidates, but also the electoral participation of younger citizens. From a practical point of view this paper will describe the development, marketing efforts and the distribution as well as the use of two of these tools : the so-called "Parteienkompass" (party compass) and the "myVote"-tool - an online voting assistance tool based on an issue-matching system comparing policy preferences between voters and candidates on an individual level. We also havea look at similar tools stemming from Voting Advice Applications (VAA) in other countries in Western Europe. The paper closes with the results of an evaluation and an outlook to further developments and on-going projects in the near future in Switzerland.