Data-driven clustering : new methods and applications

Abstract : This work is concerned with the development and application of novel unsupervised learning methods, having in mind two target applications: the analysis of forensic case data and the classification of remote sensing images. First, a method based on a symbolic optimization of the inter-sample distance measure is proposed to improve the flexibility of spectral clustering algorithms, and applied to the problem of forensic case data. This distance is optimized using a loss function related to the preservation of neighborhood structure between the input space and the space of principal components, and solutions are found using genetic programming. Results are compared to a variety of state-of--the-art clustering algorithms. Subsequently, a new large-scale clustering method based on a joint optimization of feature extraction and classification is proposed and applied to various databases, including two hyperspectral remote sensing images. The algorithm makes uses of a functional model (e.g., a neural network) for clustering which is trained by stochastic gradient descent. Results indicate that such a technique can easily scale to huge databases, can avoid the so-called out-of-sample problem, and can compete with or even outperform existing clustering algorithms on both artificial data and real remote sensing images. This is verified on small databases as well as very large problems. Résumé : Ce travail de recherche porte sur le développement et l'application de méthodes d'apprentissage dites non supervisées. Les applications visées par ces méthodes sont l'analyse de données forensiques et la classification d'images hyperspectrales en télédétection. Dans un premier temps, une méthodologie de classification non supervisée fondée sur l'optimisation symbolique d'une mesure de distance inter-échantillons est proposée. Cette mesure est obtenue en optimisant une fonction de coût reliée à la préservation de la structure de voisinage d'un point entre l'espace des variables initiales et l'espace des composantes principales. Cette méthode est appliquée à l'analyse de données forensiques et comparée à un éventail de méthodes déjà existantes. En second lieu, une méthode fondée sur une optimisation conjointe des tâches de sélection de variables et de classification est implémentée dans un réseau de neurones et appliquée à diverses bases de données, dont deux images hyperspectrales. Le réseau de neurones est entraîné à l'aide d'un algorithme de gradient stochastique, ce qui rend cette technique applicable à des images de très haute résolution. Les résultats de l'application de cette dernière montrent que l'utilisation d'une telle technique permet de classifier de très grandes bases de données sans difficulté et donne des résultats avantageusement comparables aux méthodes existantes.

Counterfactuals and causal inference: Methods and principles for social research

"Most quantitative empirical analyses are motivated by the desire to estimate the causal effect of an independent variable on a dependent variable. Although the randomized experiment is the most powerful design for this task, in most social science research done outside of psychology, experimental designs are infeasible. (Winship & Morgan, 1999, p. 659)." This quote from earlier work by Winship and Morgan, which was instrumental in setting the groundwork for their book, captures the essence of our review of Morgan and Winship's book: It is about causality in nonexperimental settings.

Changes in insulin secretion and glucose metabolism induced by dexamethasone in lean and obese females

Résumé But: Chez les individus sveltes et en bonne santé, les modifications de la sensibilité à l'insuline secondaires à l'administration de dexaméthasone pendant deux jours sont compensées par une modification de la sécrétion d'insuline, permettant le maintien de l'homéostasie glucidique. Cette étude évalue les modifications du métabolisme glucidique et de la sécrétion d'insuline induites par une administration limitée de dexaméthasone chez les femmes obèses. Méthode de recherche: Onze femmes obèses ayant une tolérance au glucose normale ont été étudiées à deux reprises, 1° sans dexaméthasone et 2° après deux jours d'administration de dexaméthasone à faible dose. Un clamp hyperglycémique comportant deux plateaux (taux plasmatique de glucose à 7.5, respectivement 10 mM) avec du glucose marqué (6.6 ²H2 glc) a été utilisé pour déterminer la sécrétion d'insuline et le métabolisme du glucose du corps entier. Les résultats ont été comparés à ceux d'un groupe de huit femmes sveltes. Résultats : Sans dexaméthasone, les femmes obèses avaient un taux d'insuline plasmatique supérieur à jeun, durant le premier pic de sécrétion d'insuline, et aux deux plateaux hyperglycémiques. Elles avaient toutefois un métabolisme glucidique normal comparé à celui des femmes sveltes, ce qui indique une compensation adéquate. Après administration de la dexaméthasone, les femmes obèses avaient une augmentation du taux d'insuline plasmatique de 66 à 92%, mais une baisse de stockage du glucose de 15.4%. Ceci contrastait avec l'augmentation du taux d'insuline plasmatique de 91 à 113% chez les femmes sveltes et l'absence de changement de stockage du glucose du corps entier. Discussion : L'administration de dexaméthasone conduit à une baisse significative du stockage du glucose du corps entier pour une glycémie fixée chez les femmes obèses mais non chez les femmes sveltes. Ceci indique que les femmes obèses sont incapables d'accroître adéquatement leur sécrétion d'insuline. Abstract: Objective: In healthy lean individuals, changes in insulin sensitivity occurring as a consequence of a 2-day dexamethasone administration are compensated for by changes in insulin secretion, allowing glucose homeostasis to be maintained. This study evaluated the changes in glucose metabolism and insulin secretion induced by short-term dexamethasone administration in obese women. Research Methods and Procedures: Eleven obese women with normal glucose tolerance were studied on two occasions, without and after 2 days of low-dose dexamethasone administration. A two-step hyperglycemic clamp (7.5 and 10 mr1/1 glucose) with 6,6 2H2 glucose was used to assess insulin secretion and whole body glucose metabolism. Results were compared with those obtained in a group of eight lean women. Results: Without dexamethasone, obese women had higher plasma insulin concentrations in the fasting state, during the first phase of insulin secretion, and at the two hyperglycemic plateaus. However, they had normal whole body glucose metabolism compared with lean women, indicating adequate compensation. After dexamethasone, obese women had a 66% to 92% increase in plasma insulin concentrations but a 15.4% decrease in whole body glucose disposal. This contrasted with lean women, who had a 91% to 113% increase in plasma insulin concentrations, with no change in whole body glucose disposal. Discussion: Dexamethasone administration led to a significant reduction in whole body glucose disposal at fixed glycemia in obese but not lean women. This indicates that obese women are unable to increase their insulin secretion appropriately.

How important are interview methods and questionnaire designs in research on self-reported juvenile delinquency? An experimental comparison of internet vs. paper-and-pencil questionnaires and different definitions of the reference period

There has been relatively little change over recent decades in the methods used in research on self-reported delinquency. Face-to-face interviews and selfadministered interviews in the classroom are still the predominant alternatives envisaged. New methods have been brought into the picture by recent computer technology, the Internet, and an increasing availability of computer equipment and Internet access in schools. In the autumn of 2004, a controlled experiment was conducted with 1,203 students in Lausanne (Switzerland), where "paper-and-pencil" questionnaires were compared with computer-assisted interviews through the Internet. The experiment included a test of two different definitions of the (same) reference period. After the introductory question ("Did you ever..."), students were asked how many times they had done it (or experienced it), if ever, "over the last 12 months" or "since the October 2003 vacation". Few significant differences were found between the results obtained by the two methods and for the two definitions of the reference period, in the answers concerning victimisation, self-reported delinquency, drug use, failure to respond (missing data). Students were found to be more motivated to respond through the Internet, take less time for filling out the questionnaire, and were apparently more confident of privacy, while the school principals were less reluctant to allow classes to be interviewed through the Internet. The Internet method also involves considerable cost reductions, which is a critical advantage if self-reported delinquency surveys are to become a routinely applied method of evaluation, particularly so in countries with limited resources. On balance, the Internet may be instrumental in making research on self-reported delinquency far more feasible in situations where limited resources so far have prevented its implementation.

Prevalence and determinants of obesity in children in public schools of Sintra, Portugal

OBJECTIVE: To assess the prevalence of and the factors related to overweight and obesity in a sample of children from the region of Sintra, Portugal. METHODS AND PROCEDURES: Cross-sectional study, stratified for freguesia with random selection of schools. Height, weight, triceps skinfold, upper arm and waist circumferences were measured, and overweight/obesity defined according to international criteria. Breast-feeding, number of daily meals and parents' height and weight data were also collected. RESULTS: One thousand two hundred and twenty-five children aged 6-10 years were assessed. Overall prevalence of overweight and obesity was 35.6% (23% overweight and 12.6% obesity). Overweight or obese children had higher triceps skinfold, upper arm circumference, arm muscle area, and waist circumference than their normal weight counterparts (P < 0.001). On multivariate analysis, relatively to a child without obese progenitors, a child with one obese progenitor had an obesity risk multiplied by 2.78 (95% confidence interval (CI): 1.76-4.38), while a child with two obese progenitors had a risk multiplied by 6.47 (95% CI: 5.59-16.19). Conversely, being picky was significantly related with a smaller risk of obesity: for boys, odds ratio (OR) = 0.15 (95% CI: 0.04-0.63); for girls, OR = 0.19 (95% CI: 0.06-0.64). Finally, no relationships were found between obesity, birth weight, birth height or breast-feeding. DISCUSSION: Prevalence of overweight and obesity are elevated among children of the Sintra region in Portugal compared to most other regions of Europe. The relationship with the parents' nutritional state stresses the need to target families for preventing obesity.

Manual reaction times and brain dynamics after 'awake surgery' of slow-growing tumors invading the parietal area. A case report

Primary objectives: Awake surgeries of slow-growing tumours invading the brain and guided by direct electrical stimulation induce major brain reorganizations accompanied with slight impairments post-operatively. In most cases, these deficits are so slight after a few days that they are often not detectable on classical neuropsychological evaluations. Consequently, this study investigated whether simple visuo-manual reaction time paradigms would sign some level of functional asymmetries between both hemispheres. Importantly, the visual stimulus was located in the saggital plane in order to limit attentional biases and to focus mainly on the inter-hemispheric asymmetry. Methods and procedures: Three patients (aged 41, 59 and 59 years) after resections in parietal regions and a control group (age¼44, SD¼6.9) were compared during simple uni- and bimanual reaction times (RTs). Main outcomes and results: Longer RTs were observed for the contralesional compared to the ipsilesional hand in the unimanual condition. This asymmetry was reversed for the bimanual condition despite longer RTs. Conclusion and clinical implications: Reaction time paradigms are useful in these patients to monitor more precisely their functional deficits, especially their level of functional asymmetry, and to understand brain (re)organization following slowgrowing lesions.

Variability of voriconazole plasma levels measured by new high-performance liquid chromatography and bioassay methods, [and] Voriconazole therapeutic drug monitoring in patients with invasive mycoses improves efficacy and safety outcomes

Rapport de synthèse1. Partie de laboratoireCette première étude décrit le développement et la validation, selon les standards internationaux, de deux techniques de mesure des concentrations sanguines de voriconazole, un nouvel agent antifongique à large spectre: 1) la chromatographic en phase liquide à haute pression et 2) le bio-essai utilisant une souche mutante de Candida hypersensible au voriconazole. Ce travail a aussi permis de mettre en évidence une importante et imprévisible variabilité inter- et intra-individuelle des concentrations sanguines de voriconazole malgré l'utilisation des doses recommandées par le fabriquant. Ce travail a été publié dans un journal avec "peer-review": "Variability of voriconazole plasma levels measured by new high- performance liquid chromatography and bioassay methods" by A. Pascual, V. Nieth, T. Calandra, J. Bille, S. Bolay, L.A. Decosterd, T. Buclin, P.A. Majcherczyk, D. Sanglard, 0. Marchetti. Antimicrobial Agents Chemotherapy, 2007; 51:137-432. Partie CliniqueCette deuxième étude a évalué de façon prospective l'impact clinique des concentrations sanguines de voriconazole sur l'efficacité et sécurité thérapeutique chez des patients atteints d'infections fongiques. Des concentrations sanguines élevées étaient significativement associés à la survenue d'une toxicité neurologique (encéphalopathie avec confusion, hallucinations et myoclonies) et des concentrations sanguines basses à une réponse insuffisante au traitement antifongique (persistance ou progression des signes cliniques et radiologiques de l'infection). Dans la majorité des cas, un ajustement de la dose de voriconazole, sur la base des concentrations mesurées, a abouti à une récupération neurologique complète ou à une résolution de l'infection, respectivement. Ce travail a été publié dans un journal avec "peer-review": " Voriconazole Therapeutic Drug Monitoring in Patients with Invasive Mycoses Improves Efficacy and Safety Outcomes" by A. Pascual, T. Calandra, S. Bolay, T. Buclin, J. Bille, and O. Marchetti. Clinical Infectious Diseases, 2008 January 15; 46(2): 201-11.Ces deux études, financées de façon conjointe par un "grant" international de la Société suisse d'infectiologie et la Société internationale de maladies infectieuses et par la Fondation pour le progrès en microbiologie médicale et maladies infectieuses (FAMMID, Lausanne), ont été réalisées au sein du Service des Maladies Infectieuses, Département de Médecine, au CHUV, en étroite collaboration avec la Division de Pharmacologie Clinique, Département de Médecine, au CHUV et l'Institut de Microbiologie du CHUV et de l'Université de Lausanne.

Expression and secretion of the atrial natriuretic peptide in human adipose tissue and preadipocytes.

OBJECTIVE: Atrial natriuretic peptide (ANP) is a secretory hormone displaying diuretic, natriuretic, and vasorelaxant activities. Recently, its lipolytic activity has been reported. Since the expression of ANP in adipose tissue has not been documented, we used real-time reverse transcriptase polymerase chain reaction (RT-PCR) to investigate the expression of ANP in human adipose tissue and preadipocytes. RESEARCH METHODS AND PROCEDURES: RNA was extracted from the human adipose tissue of severely obese premenopausal women as well as from human preadipocytes. For human preadipocytes, two cell systems were investigated: the human preadipose immortalized (Chub-S7) cells, a well-characterized human preadipose cell line, and primary preadipocytes derived from the stromal vascular fraction of the human adipose tissue. We measured the mRNA of ANP, of corin (a transmembrane serine protease involved in the conversion of pro-ANP to ANP) and of uncoupling protein 2 (UCP2; a control gene known to be ubiquitously expressed). The expression of ANP was also investigated using immunofluorescence and radioimmunoassay in Chub-S7 cells and human primary preadipocytes in culture. RESULTS: Our results indicate that ANP and corin are expressed at the mRNA level in human adipose tissue and preadipocytes. Immunofluorescence experiments demonstrated that pro-ANP was expressed in Chub-S7 cells. In addition, ANP secretion could be measured in Chub-S7 cells and human primary preadipocytes in culture. Rosiglitazone, a selective peroxisome proliferator-activated receptor type gamma (PPAR-gamma) agonist promoting adipocyte differentiation, was found to modulate both ANP expression and secretion in preadipocytes. DISCUSSION: Our findings suggest the existence of an autocrine/paracrine system for ANP in the human adipose tissue whose implications in lipolysis and cardiovascular function need to be further explored.

Overweight in Swiss children and associations with children's and parents' characteristics et Worldwide trends in childhood obesity

(1) Surpoids chez les enfants suisses et associations avec certaines caractéristiques chez les enfants et leurs parents Le but de cette étude était de mesurer la prévalence du surpoids et de l?obésité chez les enfants de sixième année du canton de Vaud (âge moyen de 12 ans) et les facteurs associés au surpoids. Les données ont été récoltées lors d?une étude menée par l?Institut universitaire de médecine sociale et préventive. Tous les enfants scolarisés en 6ème année à l?école publique du canton de Vaud entre septembre 2005 et mai 2006 étaient éligibles à participer à cette étude. Le taux de participation a atteint 76% (soit 5207 enfants de 12,3 ans en moyenne). Le poids et la taille des enfants ont été mesurés à l?école par des assistants de recherche et les enfants ont rempli, en classe, un questionnaire structuré sur leur mode de vie (notamment : temps quotidien passé à regarder la télévision, à jouer à des jeux sur écran; fréquence de la pratique de diverses activités physiques; fréquence de la consommation de fruits ou de légumes). Des informations sur les parents (niveau d?éducation, nationalité, poids et taille) ont été récoltées au moyen d?un questionnaire structuré envoyé par courrier à ceux-ci. Nous avons utilisé les critères de l?International Obesity Task Force, qui définit les valeurs-seuils de l'indice de masse corporelle pour le surpoids et pour l?obésité, par age et par sexe. La prévalence du surpoids (obésité incluse) dans la population était de 15% chez les garçons et de 12% chez les filles, et la prévalence de l?obésité était de 2% dans les deux sexes. Nous avons trouvé que le surpoids était associé de façon indépendante avec le temps passé à regarder la télévision, ainsi qu?avec certaines caractéristiques des parents, comme le surpoids, un bas niveau d?éducation et une nationalité étrangère. En conclusion, un enfant sur sept est en surpoids ou obèse dans le canton de Vaud. Ces chiffres indiquent un important défi de santé publique, même si cette prévalence dans le canton de Vaud est, actuellement, moindre que dans beaucoup d?autres pays d?Europe, et bien moindre qu?en Amérique du Nord. Les associations entre le surpoids infantile et le temps passé à regarder la télévision, ainsi que les associations avec des variables liées au milieu socio-culturel des parents indiquent plusieurs pistes d?intervention pour prévenir le surpoids chez les enfants. Il est probable que les mesures de prévention ne devraient pas se limiter aux approches individuelles, mais devraient aussi inclure des mesures structurelles sur l?environnement social, physique et économique visant à réduire les facteurs obésogènes dans la société.<br/><br/>Objective: The objective was to assess the prevalence of overweight and obesity in children in a canton of Switzerland and the association with various characteristics of the parents and the children. Research Methods and Procedures: A cross-sectional survey was conducted in all children of the sixth school grade of the canton of Vaud, Switzerland. Weight and height were measured, and selected lifestyle variables were assessed with a self-administered semiquantitative questionnaire. Information on children?s parents was gathered through a mailed structured questionnaire. Overweight and obesity were based on the International Obesity Task Force criteria. Results: Of 6873 eligible children, 5207 (76%) participated (2621 boys, 2586 girls; mean age, 12.3 years; standard deviation, 0.5 year). The prevalence of overweight (including obesity) was 15.0% (95% confidence interval, 13.7% to 16.4%) in boys and 12.4% (11.1% to 13.7%) in girls, and the prevalence of obesity was 1.8% (1.3% to 2.3%) and 1.7% (1.2% to 2.2%), respectively. In both univariate and multivariate analyses, overweight was strongly associated with high television viewing time and selected characteristics of the parents (overweight, low educational level, and foreign nationality). Discussion: The prevalence of pediatric overweight and obesity was lower in this region of Switzerland than in several European countries. The correlates of overweight found in this region suggest areas for potential interventions.

National Survey of Noncommunicable Diseases in Seychelles, 2013-2014 (Seychelles Heart Study IV) : methods and main findings

This report provides information on selected summary results of the National Survey of Noncommunicable Diseases in Seychelles in 2013‐2014 (Seychelles Heart Study IV). The survey is also referred shortly as the "2013 Survey" in this report. Overall crude results were reported in a comprehensive report in November 2014. Further detailed analyses and recommendations on particular topics will be performed separately.