Abdominal infections in the intensive care unit: characteristics, treatment and determinants of outcome.

BACKGROUND: Abdominal infections are frequent causes of sepsis and septic shock in the intensive care unit (ICU) and are associated with adverse outcomes. We analyzed the characteristics, treatments and outcome of ICU patients with abdominal infections using data extracted from a one-day point prevalence study, the Extended Prevalence of Infection in the ICU (EPIC) II. METHODS: EPIC II included 13,796 adult patients from 1,265 ICUs in 75 countries. Infection was defined using the International Sepsis Forum criteria. Microbiological analyses were performed locally. Participating ICUs provided patient follow-up until hospital discharge or for 60 days. RESULTS: Of the 7,087 infected patients, 1,392 (19.6%) had an abdominal infection on the study day (60% male, mean age 62 ± 16 years, SAPS II score 39 ± 16, SOFA score 7.6 ± 4.6). Microbiological cultures were positive in 931 (67%) patients, most commonly Gram-negative bacteria (48.0%). Antibiotics were administered to 1366 (98.1%) patients. Patients who had been in the ICU for ≤ 2 days prior to the study day had more Escherichia coli, methicillin-sensitive Staphylococcus aureus and anaerobic isolates, and fewer enterococci than patients who had been in the ICU longer. ICU and hospital mortality rates were 29.4% and 36.3%, respectively. ICU mortality was higher in patients with abdominal infections than in those with other infections (29.4% vs. 24.4%, p < 0.001). In multivariable analysis, hematological malignancy, mechanical ventilation, cirrhosis, need for renal replacement therapy and SAPS II score were independently associated with increased mortality. CONCLUSIONS: The characteristics, microbiology and antibiotic treatment of abdominal infections in critically ill patients are diverse. Mortality in patients with isolated abdominal infections was higher than in those who had other infections.

Affective and physiological responses to environmental noises and music

Research suggests that respiratory patterns may reflect general dimensions of emotional response. In this study, we investigated the relationships between judgments of affective valence (pleasantness) and arousal and respiratory responses to acoustic stimuli. Sixteen environmental noises and 16 musical fragments of 30 s duration were presented to 31 participants, while respiration, skin conductance level and heart rate were recorded. Judgments of valence and arousal were registered using the 9-point Self-Assessment Manikin. For noises, breathing accelerated and minute ventilation augmented with decreases in pleasantness for low-arousal stimuli and with increases in arousal for positive stimuli. For music, breathing accelerated and minute ventilation augmented with increases both in rated valence and arousal. Skin conductance level increased with arousal ratings for music but not for noises, whereas mean heart rate increased with rated arousal for noises but not for music. Although both noises and music are sound-vibrations, differences in the relationships between affective judgments and physiological responses were found suggesting differences in the processing of the two types of acoustic stimuli. [Authors]

Pneumocystis jirovecii genotype associated with increased death rate of HIV-infected patients with pneumonia.

Pneumocystis jirovecii dihydropteroate synthase (DHPS) mutations have been associated with failure of sulfa prophylaxis; their effect on the outcome of patients with P. jirovecii pneumonia (PCP) remains controversial. P. jirovecii DHPS polymorphisms and genotypes were identified in 112 cases of PCP in 110 HIV-infected patients by using PCR single-strand conformation polymorphism. Of the 110 patients observed, 21 died; 18 of those deaths were attributed to PCP. Thirty-three percent of the PCP cases involved a P. jirovecii strain that had 1 or both DHPS mutations. The presence or absence of DHPS mutations had no effect on the PCP mortality rate within 1 month, whereas P.jirovecii type 7 and mechanical ventilation at PCP diagnosis were associated with an increased risk of death caused by PCP. Mechanical ventilation at PCP diagnosis was also associated with an increased risk of sulfa treatment failure at 5 days.

Post-exercise parasympathetic reactivation and sensibility to hypoxia

Introduction Exposure to hypoxia leads to several reactions of the organism, which try to compensate the reduced oxygen level in the blood. Acute response is characterized by an increase in pulmonary ventilation (Hypoxia Ventilatory Response, HVR) and in cardiac output (cardiac response to hypoxia). Heart rate (HR) at rest and during exercise is higher at high altitude than at sea level, whereas HRmax is lower. These cardiac adaptations are partially explained by an increased sympathetic stimulation associated with a reduced parasympathetic tone (12). The precise mechanisms of HRmax decline in acute hypoxia are however still to be identified, although several hypothesis have been suggested, such as a direct effect of hypoxia on the electrophysiological properties, an influence of skeletal maximal VO2 or a modulation of the autonomic nervous system (8). Some authors have reported that endurance trained athletes present an increased sensitivity to hypoxia shown by a large reduction in VO2max and an important decrease in arterial saturation. (9,11, 13) A hypoxia test can assess the sensibility of chemoreceptors to the reduction of oxygen by calculating hypoxic ventilatory and cardiac responses, knowing that low sensibility is correlated with poor acclimatization. Two parameters results from the differences in ventilation (and heart rate) divided by the difference in the arterial oxygen saturation between normoxia and hypoxia (18). Objective The hypothesis tested by this study is that parasympathetic reactivation after moderate effort in hypoxic condition can be used as a marker of individual sensibility to hypoxia. Parasympathetic reactivation is a marker of vagal tone that predict endurance capacity and aerobic fitness (2,7). Methods Subjects This study uses data obtained from two groups of athletes participating into two larger studies about adaptation to hypoxia. One group is composed of elite athletes (Swiss ski mountaineering team), the other one of mid-level athletes (ski mountaineering amateurs). The particularity of this target population is that they often train at high altitude, and therefore could show a better response to hypoxia than athleltes of other disciplines. Protocol The athletes performed a submaximal exercise (6min run at 9 km/h, flat) followed by 10 min of seated rest either in an hypoxic chamber (simulated altitude of 3000m) or in normoxic conditions. During the resting phase parasympathetic reactivation was assessed by beat-to-beat HR measurements.A test of tolerance to altitude was also performed. Analysis Parasympathetic reactivation, assessed by the calculation of the root mean square of successive differences in the R-R intervals (RMSSD)(4), is compared to individual responses at altitude, in order to appreciate the correlation between the two phenomena.

Contribution of fine particulate matter sources to indoor exposure in bars, restaurants, and cafes

This study investigated the contribution of sources and establishment characteristics, on the exposure to fine particulate matter (PM(2.5)) in the non-smoking sections of bars, cafes, and restaurants in central Zurich. PM(2.5)-exposure was determined with a nephelometer. A random sample of hospitality establishments was investigated on all weekdays, from morning until midnight. Each visit lasted 30 min. Numbers of smokers and other sources, such as candles and cooking processes, were recorded, as were seats, open windows, and open doors. Ambient air pollution data were obtained from public authorities. Data were analysed using robust MM regression. Over 14 warm, sunny days, 102 establishments were measured. Average establishment PM(2.5) concentrations were 64.7 microg/m(3) (s.d. = 73.2 microg/m(3), 30-min maximum 452.2 microg/m(3)). PM(2.5) was significantly associated with the number of smokers, percentage of seats occupied by smokers, and outdoor PM. Each smoker increased PM(2.5) on average by 15 microg/m(3). No associations were found with other sources, open doors or open windows. Bars had more smoking guests and showed significantly higher concentrations than restaurants and cafes. Smokers were the most important PM(2.5)-source in hospitality establishments, while outdoor PM defined the baseline. Concentrations are expected to be even higher during colder, unpleasant times of the year. PRACTICAL IMPLICATIONS: Smokers and ambient air pollution are the most important sources of fine airborne particulate matter (PM(2.5)) in the non-smoking sections of bars, restaurants, and cafes. Other sources do not significantly contribute to PM(2.5)-levels, while opening doors and windows is not an efficient means of removing pollutants. First, this demonstrates the impact that even a few smokers can have in affecting particle levels. Second, it implies that creating non-smoking sections, and using natural ventilation, is not sufficient to bring PM(2.5) to levels that imply no harm for employees and non-smoking clients. [Authors]

Interhospital aeromedical transfers: need for a specific training

Introduction: The interhospital aeromedical transfers concern mainly unstable patients presenting with multiple organ failure and among whom the intensity, as well as the difficulty of care are often underestimated. Material and method: Retrospective analysis of 2094 interhospital transfers made by the Rega helicopter of Lausanne between 2003 and 2008, with description of the pathologies, the demographic characteristics, as well as the medical and technical difficulties. Results: Male patient are overrepresented (65 vs 35%), their age being 40-80 years old. On the other hand, we note an important number (10%) of patients less than 10 years. The average time of flight is 13 minutes, with a high rate of night transfers (34%). In 73% of the cases, the transfers concern medical, especially cardiovascular and neurological situations. Trauma and surgical situations account for less than 20%. The transferred patients often require mechanical ventilation (27%), as well as invasive measure of blood pressure (11%), particularly in cases of neurological, pulmonary or cardiac diseases. In 6% of the cases, we note a haemodynamic instability, requiring the use of catecholamines. In 1% a cardiopulmonary resuscitation was initiated during the flight. The overall mortality at 48 hours is about 5%. Conclusions: In spite of relatively brief flights, the complexity and the variety of the pathologies require a specific training for the medical teams involved in the transfers, concerning the aspects of resuscitation, mechanical ventilation, and the knowledge of advanced life support in paediatrics, traumatology and cardiopulmonary intensive care.

Surréanimation liquidienne pré et intrahospitatière des patients brûlés: fréquente et néfaste [Pre- and intra-hospital over-resuscitation in burns: frequent and deleterious].

Major burns are characterized by an initial capillary leak which requires fluid resuscitation for hemodynamic stabilisation. While under-resuscitation was the major cause of death until the 80ies, over-resuscitation has become an important source of complications: abdominal compartment syndrome, escharotomies, impaired gas exchange and prolonged mechanical ventilation and hospital stay. The fluid creep started in the 90ies with an increasing proportion of the first 24 hours' fluid delivery above the 4 ml/kg/% BSA Parkland prediction. The first alerts were published under the form of case reports of increased mortality due to abdominal compartment syndrome and respiratory failure. The paper analyses the causes of this fluid creep, and the ways to prevent it, which includes rationing prehospital fluid delivery, avoiding early colloids and permissive hypovolemia.

Respiratory distress in a one-month-old child suffering brachial plexus palsy.

This paper describes a one-month-old girl presenting with respiratory and growth failure due to diaphragmatic paralysis associated with left brachial plexus palsy after forceps delivery. Despite continuous positive pressure ventilation and nasogastric feeding, the situation did not improve and a laparoscopic diaphragmatic plication had to be performed. When dealing with a child born with brachial plexus palsy, one must think of this possible association and if necessary proceed to the complementary radiological examinations. The treatment must avoid complications like feeding difficulties and failure to thrive, respiratory infections or atelectasis. It includes intensive support and a good evaluation of the prognosis of the lesion to decide the best moment for a surgical therapy.

A randomized trial of glutamine and antioxidants in critically ill patients.

BACKGROUND: Critically ill patients have considerable oxidative stress. Glutamine and antioxidant supplementation may offer therapeutic benefit, although current data are conflicting. METHODS: In this blinded 2-by-2 factorial trial, we randomly assigned 1223 critically ill adults in 40 intensive care units (ICUs) in Canada, the United States, and Europe who had multiorgan failure and were receiving mechanical ventilation to receive supplements of glutamine, antioxidants, both, or placebo. Supplements were started within 24 hours after admission to the ICU and were provided both intravenously and enterally. The primary outcome was 28-day mortality. Because of the interim-analysis plan, a P value of less than 0.044 at the final analysis was considered to indicate statistical significance. RESULTS: There was a trend toward increased mortality at 28 days among patients who received glutamine as compared with those who did not receive glutamine (32.4% vs. 27.2%; adjusted odds ratio, 1.28; 95% confidence interval [CI], 1.00 to 1.64; P=0.05). In-hospital mortality and mortality at 6 months were significantly higher among those who received glutamine than among those who did not. Glutamine had no effect on rates of organ failure or infectious complications. Antioxidants had no effect on 28-day mortality (30.8%, vs. 28.8% with no antioxidants; adjusted odds ratio, 1.09; 95% CI, 0.86 to 1.40; P=0.48) or any other secondary end point. There were no differences among the groups with respect to serious adverse events (P=0.83). CONCLUSIONS: Early provision of glutamine or antioxidants did not improve clinical outcomes, and glutamine was associated with an increase in mortality among critically ill patients with multiorgan failure. (Funded by the Canadian Institutes of Health Research; ClinicalTrials.gov number, NCT00133978.).

Airflow decline after myeloablative allogeneic hematopoietic cell transplantation: the role of community respiratory viruses.

We conducted a 12-year retrospective study to determine the effects that the community respiratory-virus species and the localization of respiratory-tract virus infection have on severe airflow decline, a serious and fatal complication occurring after hematopoietic cell transplantation (HCT). Of 132 HCT recipients with respiratory-tract virus infection during the initial 100 days after HCT, 50 (38%) developed airflow decline < or =1 year after HCT. Lower-respiratory-tract infection with parainfluenza (odds ratio [OR], 17.9 [95% confidence interval {CI}, 2.0-160]; P=.01) and respiratory syncytial virus (OR, 3.6 [95% CI, 1.0-13]; P=.05) independently increased the risk of development of airflow decline < or =1 year after HCT. The airflow decline was immediately detectable after infection and was strongest for lower-respiratory-tract infection with parainfluenza virus; it stabilized during the months after the respiratory-tract virus infection, but, at < or =1 year after HCT, the initial lung function was not restored. Thus, community respiratory virus-associated airflow decline seems to be specific to viral species and infection localization.

Airborne exposures to monoethanolamine, glycol ethers, and benzyl alcohol during professional cleaning: a pilot study

A growing body of epidemiologic evidence suggests an association between exposure to cleaning products and respiratory dysfunction. Due to the lack of quantitative assessments of respiratory exposures to airborne irritants and sensitizers among professional cleaners, the culpable substances have yet to be identified.Purpose: Focusing on previously identified irritants, our aims were to determine (i) airborne concentrations of monoethanolamine (MEA), glycol ethers, and benzyl alcohol (BA) during different cleaning tasks performed by professional cleaning workers and assess their determinants; and (ii) air concentrations of formaldehyde, a known indoor air contaminant. METHODS: Personal air samples were collected in 12 cleaning companies, and analyzed by conventional methods. RESULTS: Nearly all air concentrations [MEA (n = 68), glycol ethers (n = 79), BA (n = 15), and formaldehyde (n = 45)] were far below (<1/10) of the corresponding Swiss occupational exposure limits (OEL), except for ethylene glycol mono-n-butyl ether (EGBE). For butoxypropanol and BA, no OELs exist. Although only detected once, EGBE air concentrations (n = 4) were high (49.48-58.72mg m(-3)), and close to the Swiss OEL (49mg m(-3)). When substances were not noted as present in safety data sheets of cleaning products used but were measured, air concentrations showed no presence of MEA, while the glycol ethers were often present, and formaldehyde was universally detected. Exposure to MEA was affected by its amount used (P = 0.036), and spraying (P = 0.000) and exposure to butoxypropanol was affected by spraying (P = 0.007) and cross-ventilation (P = 0.000). CONCLUSIONS: Professional cleaners were found to be exposed to multiple airborne irritants at low concentrations, thus these substances should be considered in investigations of respiratory dysfunctions in the cleaning industry; especially in specialized cleaning tasks such as intensive floor cleaning.

Impact of lung function changes after induction radiochemotherapy on resected T4 non-small cell lung cancer outcome.

BACKGROUND: Induction radiochemotherapy, followed by resection, for T4 non-small cell lung cancer, has shown promising long-term survival but may be associated with increased postoperative morbidity and death, depending on patient selection. Here, we determined the effect of induction radiochemotherapy on pulmonary function and whether postinduction pulmonary function changes predict hospital morbidity and death and long-term survival. METHODS: A consecutive prospective cohort of 72 patients with T4 N0-2 M0 non-small cell lung cancer managed by radiochemotherapy, followed by resection, is reported. All patients underwent thoracoabdominal computed tomography or fusion positron emission tomography-computed tomography, brain imaging, mediastinoscopy, echocardiography, ventilation-perfusion scintigraphy, and pulmonary function testing before and after induction therapy. Resection was performed if the postoperative forced expiratory volume in 1 second and diffusion capacity of the lung for carbon monoxide exceeded 30% predicted and if the postoperative maximum oxygen consumption exceeded 10 mL/kg/min. RESULTS: The postoperative 90-day mortality rate was 8% (lobectomy, 2%; pneumonectomy, 21%; p=0.01). All deaths after pneumonectomy occurred after right-sided procedures. The 3-year and 5-year survival was 50% (95% confidence interval, 36% to 62%) and 45% (95% confidence interval, 31% to 57%) and was significantly associated with completeness of resection (p=0.004) and resection type (pneumonectomy vs lobectomy, p=0.01). There was no correlation between postinduction pulmonary function changes and postoperative morbidity or death or long-term survival in patients managed by lobectomy or pneumonectomy. CONCLUSIONS: In properly selected patients with T4 N0-2 M0 non-small cell lung cancer, resection after induction radiochemotherapy can be performed with a reasonable postoperative mortality rate and long-term survival, provided the resection is complete and a right-sided pneumonectomy is avoided. Postinduction pulmonary function changes did not correlate with postoperative morbidity or death or with long-term outcome.

Overall Burden of Healthcare-Associated Infections Among Surgical Patients: Results of a National Study.

OBJECTIVE:: To assess the overall burden of healthcare-associated infections (HAIs) in patients exposed and nonexposed to surgery. BACKGROUND:: Targeted HAI surveillance is common in healthcare institutions, but may underestimate the overall burden of disease. METHODS:: Prevalence study among patients hospitalized in 50 acute care hospitals participating in the Swiss Nosocomial Infection Prevalence surveillance program. RESULTS:: Of 8273 patients, 3377 (40.8%) had recent surgery. Overall, HAI was present in 358 (10.6%) patients exposed to surgery, but only in 206 (4.2%) of 4896 nonexposed (P < 0.001). Prevalence of surgical site infection (SSI) was 5.4%. Healthcare-associated infections prevalence excluding SSI was 6.5% in patients with surgery and 4.7% in those without (P < 0.0001). Patients exposed to surgery carried less intrinsic risk factors for infection (age >60 years, 55.6% vs 63.0%; American Society of Anesthesiologists score >3, 5.9% vs 9.3%; McCabe for rapidly fatal disease, 3.9% vs 6.6%; Charlson comorbidity index >2, 12.3% vs 20.9%, respectively; all P < 0.001) than those nonexposed, but more extrinsic risk factors (urinary catheters, 39.6% vs 14.1%; central venous catheters, 17.8% vs 7.1%; mechanical ventilation, 4.7% vs 1.3%; intensive care stay, 18.3% vs 8.8%, respectively; all P < 0.001). Exposure to surgery independently predicted an increased risk of HAI (odds ratio 2.43; 95% CI 2.0-3.0). CONCLUSIONS:: Despite a lower intrinsic risk, patients exposed to surgery carried more than twice the overall HAI burden than those nonexposed; almost half was accountable to SSI. Extending infection control efforts beyond SSI prevention in these patients might be rewarding, especially because of the extrinsic nature of risk factors.

Exclusion of pulmonary embolism using C-reactive protein and D-dimer. A prospective comparison.

Our goal was to evaluate the diagnostic utility of C-reactive protein (CRP) alone or combined with clinical probability assessment in patients with suspected pulmonary embolism (PE), and to compare its performance to a D-dimer assay. We conducted a prospective study in which we performed a common immuno-turbidimetric CRP test and a rapid enzyme-linked immunosorbent assay (ELISA) D-dimer test in 259 consecutive outpatients with suspected PE at the emergency department of a teaching hospital. We assessed clinical probability of PE by a validated prediction rule overridden by clinical judgment. Patients with D-dimer levels &gt; or = 500 microg/l underwent a work-up consisting of lower-limb venous ultrasound, spiral computerized tomography, ventilation-perfusion scan, or pulmonary angiography. Patients were followed up for three months. Seventy-seven (30%) of the patients had PE. The CRP alone had a sensitivity of 84% (95% confidence interval [CI).: 74 to 92%) and a negative predictive value (NPV) of 87% (95% CI: 78 to 93%) at a cutpoint of 5 mg/l. Overall, 61 (24%) patients with a low clinical probability of PE had a CRP &lt; 5 mg/l. Due to the low prevalence of PE (9%) in this subgroup, the NPV increased to 97% (95% CI: 89 to 100%). The D-dimer (cutpoint 500 micro g/l) showed a sensitivity of 100% (95% CI: 95 to 100%) and a NPV of 100% (95% CI: 94 to 100%) irrespective of clinical probability and accurately rule out PE in 56 (22%) patients. Standard CRP tests alone or combined with clinical probability assessment cannot safely exclude PE.

Machine learning for geospatial data : algorithms, software tools and case studies

Résumé Cette thèse est consacrée à l'analyse, la modélisation et la visualisation de données environnementales à référence spatiale à l'aide d'algorithmes d'apprentissage automatique (Machine Learning). L'apprentissage automatique peut être considéré au sens large comme une sous-catégorie de l'intelligence artificielle qui concerne particulièrement le développement de techniques et d'algorithmes permettant à une machine d'apprendre à partir de données. Dans cette thèse, les algorithmes d'apprentissage automatique sont adaptés pour être appliqués à des données environnementales et à la prédiction spatiale. Pourquoi l'apprentissage automatique ? Parce que la majorité des algorithmes d'apprentissage automatiques sont universels, adaptatifs, non-linéaires, robustes et efficaces pour la modélisation. Ils peuvent résoudre des problèmes de classification, de régression et de modélisation de densité de probabilités dans des espaces à haute dimension, composés de variables informatives spatialisées (« géo-features ») en plus des coordonnées géographiques. De plus, ils sont idéaux pour être implémentés en tant qu'outils d'aide à la décision pour des questions environnementales allant de la reconnaissance de pattern à la modélisation et la prédiction en passant par la cartographie automatique. Leur efficacité est comparable au modèles géostatistiques dans l'espace des coordonnées géographiques, mais ils sont indispensables pour des données à hautes dimensions incluant des géo-features. Les algorithmes d'apprentissage automatique les plus importants et les plus populaires sont présentés théoriquement et implémentés sous forme de logiciels pour les sciences environnementales. Les principaux algorithmes décrits sont le Perceptron multicouches (MultiLayer Perceptron, MLP) - l'algorithme le plus connu dans l'intelligence artificielle, le réseau de neurones de régression généralisée (General Regression Neural Networks, GRNN), le réseau de neurones probabiliste (Probabilistic Neural Networks, PNN), les cartes auto-organisées (SelfOrganized Maps, SOM), les modèles à mixture Gaussiennes (Gaussian Mixture Models, GMM), les réseaux à fonctions de base radiales (Radial Basis Functions Networks, RBF) et les réseaux à mixture de densité (Mixture Density Networks, MDN). Cette gamme d'algorithmes permet de couvrir des tâches variées telle que la classification, la régression ou l'estimation de densité de probabilité. L'analyse exploratoire des données (Exploratory Data Analysis, EDA) est le premier pas de toute analyse de données. Dans cette thèse les concepts d'analyse exploratoire de données spatiales (Exploratory Spatial Data Analysis, ESDA) sont traités selon l'approche traditionnelle de la géostatistique avec la variographie expérimentale et selon les principes de l'apprentissage automatique. La variographie expérimentale, qui étudie les relations entre pairs de points, est un outil de base pour l'analyse géostatistique de corrélations spatiales anisotropiques qui permet de détecter la présence de patterns spatiaux descriptible par une statistique. L'approche de l'apprentissage automatique pour l'ESDA est présentée à travers l'application de la méthode des k plus proches voisins qui est très simple et possède d'excellentes qualités d'interprétation et de visualisation. Une part importante de la thèse traite de sujets d'actualité comme la cartographie automatique de données spatiales. Le réseau de neurones de régression généralisée est proposé pour résoudre cette tâche efficacement. Les performances du GRNN sont démontrées par des données de Comparaison d'Interpolation Spatiale (SIC) de 2004 pour lesquelles le GRNN bat significativement toutes les autres méthodes, particulièrement lors de situations d'urgence. La thèse est composée de quatre chapitres : théorie, applications, outils logiciels et des exemples guidés. Une partie importante du travail consiste en une collection de logiciels : Machine Learning Office. Cette collection de logiciels a été développée durant les 15 dernières années et a été utilisée pour l'enseignement de nombreux cours, dont des workshops internationaux en Chine, France, Italie, Irlande et Suisse ainsi que dans des projets de recherche fondamentaux et appliqués. Les cas d'études considérés couvrent un vaste spectre de problèmes géoenvironnementaux réels à basse et haute dimensionnalité, tels que la pollution de l'air, du sol et de l'eau par des produits radioactifs et des métaux lourds, la classification de types de sols et d'unités hydrogéologiques, la cartographie des incertitudes pour l'aide à la décision et l'estimation de risques naturels (glissements de terrain, avalanches). Des outils complémentaires pour l'analyse exploratoire des données et la visualisation ont également été développés en prenant soin de créer une interface conviviale et facile à l'utilisation. Machine Learning for geospatial data: algorithms, software tools and case studies Abstract The thesis is devoted to the analysis, modeling and visualisation of spatial environmental data using machine learning algorithms. In a broad sense machine learning can be considered as a subfield of artificial intelligence. It mainly concerns with the development of techniques and algorithms that allow computers to learn from data. In this thesis machine learning algorithms are adapted to learn from spatial environmental data and to make spatial predictions. Why machine learning? In few words most of machine learning algorithms are universal, adaptive, nonlinear, robust and efficient modeling tools. They can find solutions for the classification, regression, and probability density modeling problems in high-dimensional geo-feature spaces, composed of geographical space and additional relevant spatially referenced features. They are well-suited to be implemented as predictive engines in decision support systems, for the purposes of environmental data mining including pattern recognition, modeling and predictions as well as automatic data mapping. They have competitive efficiency to the geostatistical models in low dimensional geographical spaces but are indispensable in high-dimensional geo-feature spaces. The most important and popular machine learning algorithms and models interesting for geo- and environmental sciences are presented in details: from theoretical description of the concepts to the software implementation. The main algorithms and models considered are the following: multi-layer perceptron (a workhorse of machine learning), general regression neural networks, probabilistic neural networks, self-organising (Kohonen) maps, Gaussian mixture models, radial basis functions networks, mixture density networks. This set of models covers machine learning tasks such as classification, regression, and density estimation. Exploratory data analysis (EDA) is initial and very important part of data analysis. In this thesis the concepts of exploratory spatial data analysis (ESDA) is considered using both traditional geostatistical approach such as_experimental variography and machine learning. Experimental variography is a basic tool for geostatistical analysis of anisotropic spatial correlations which helps to understand the presence of spatial patterns, at least described by two-point statistics. A machine learning approach for ESDA is presented by applying the k-nearest neighbors (k-NN) method which is simple and has very good interpretation and visualization properties. Important part of the thesis deals with a hot topic of nowadays, namely, an automatic mapping of geospatial data. General regression neural networks (GRNN) is proposed as efficient model to solve this task. Performance of the GRNN model is demonstrated on Spatial Interpolation Comparison (SIC) 2004 data where GRNN model significantly outperformed all other approaches, especially in case of emergency conditions. The thesis consists of four chapters and has the following structure: theory, applications, software tools, and how-to-do-it examples. An important part of the work is a collection of software tools - Machine Learning Office. Machine Learning Office tools were developed during last 15 years and was used both for many teaching courses, including international workshops in China, France, Italy, Ireland, Switzerland and for realizing fundamental and applied research projects. Case studies considered cover wide spectrum of the real-life low and high-dimensional geo- and environmental problems, such as air, soil and water pollution by radionuclides and heavy metals, soil types and hydro-geological units classification, decision-oriented mapping with uncertainties, natural hazards (landslides, avalanches) assessments and susceptibility mapping. Complementary tools useful for the exploratory data analysis and visualisation were developed as well. The software is user friendly and easy to use.