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.

The emergence of forensic nursing and advanced nursing practice in Switzerland: an innovative case study consultation.

BACKGROUND AND METHODS: The objectives of this article were to systematically describe and examine the novel roles and responsibilities assumed by nurses in a forensic consultation for victims of violence at a University Hospital in French-speaking Switzerland. Utilizing a case study methodology, information was collected from two main sources: (a) discussion groups with nurses and forensic pathologists and (b) a review of procedures and protocols. Following a critical content analysis, the roles and responsibilities of the forensic nurses were described and compared with the seven core competencies of advanced nursing practice as outlined by Hamric, Spross, and Hanson (2009). RESULTS: Advanced nursing practice competencies noted in the analysis included "direct clinical practice," "coaching and guidance," and "collaboration." The role of the nurse in terms of "consultation," "leadership," "ethics," and "research" was less evident in the analysis. DISCUSSION AND CONCLUSION: New forms of nursing are indeed practiced in the forensic clinical setting, and our findings suggest that nursing practice in this domain is following the footprints of an advanced nursing practice model. Further reflections are required to determine whether the role of the forensic nurse in Switzerland should be developed as a clinical nurse specialist or that of a nurse practitioner.

The Emergence of Forensic Nursing and Advanced Nursing Practice in Switzerland: An Innovative Case Study Consultation.

BACKGROUND AND METHODS:: The objectives of this article were to systematically describe and examine the novel roles and responsibilities assumed by nurses in a forensic consultation for victims of violence at a University Hospital in French-speaking Switzerland. Utilizing a case study methodology, information was collected from two main sources: (a) discussion groups with nurses and forensic pathologists and (b) a review of procedures and protocols. Following a critical content analysis, the roles and responsibilities of the forensic nurses were described and compared with the seven core competencies of advanced nursing practice as outlined by Hamric, Spross, and Hanson (2009). RESULTS:: Advanced nursing practice competencies noted in the analysis included "direct clinical practice," "coaching and guidance," and "collaboration." The role of the nurse in terms of "consultation," "leadership," "ethics," and "research" was less evident in the analysis. DISCUSSION AND CONCLUSION:: New forms of nursing are indeed practiced in the forensic clinical setting, and our findings suggest that nursing practice in this domain is following the footprints of an advanced nursing practice model. Further reflections are required to determine whether the role of the forensic nurse in Switzerland should be developed as a clinical nurse specialist or that of a nurse practitioner.

Ultra-performance liquid chromatography mass spectrometry and sensitive bioassay methods for quantification of posaconazole plasma concentrations after oral dosing.

Posaconazole (POS) is a new antifungal agent for prevention and therapy of mycoses in immunocompromised patients. Variable POS pharmacokinetics after oral dosing may influence efficacy: a trough threshold of 0.5 ?g/ml has been recently proposed. Measurement of POS plasma concentrations by complex chromatographic techniques may thus contribute to optimize prevention and management of life-threatening infections. No microbiological analytical method is available. The objective of this study was to develop and validate a new simplified ultra-performance liquid chromatography tandem mass spectrometry (UPLC-MS/MS) method and a sensitive bioassay for quantification of POS over the clinical plasma concentration range. The UPLC-MS/MS equipment consisted of a triple quadrupole mass spectrometer, an electrospray ionization (ESI) source, and a C(18) analytical column. The Candida albicans POS-hypersusceptible mutant (MIC of 0.002 ?g/ml) ?cdr1 ?cdr2 ?flu ?mdr1 ?can constructed by targeted deletion of multidrug efflux transporters and calcineurin genes was used for the bioassay. POS was extracted from plasma by protein precipitation with acetonitrile-methanol (75%/25%, vol/vol). Reproducible standard curves were obtained over the range 0.014 to 12 (UPLC-MS/MS) and 0.028 to 12 ?g/ml (bioassay). Intra- and interrun accuracy levels were 106% ± 2% and 103% ± 4% for UPLC-MS/MS and 102% ± 8% and 104% ± 1% for bioassay, respectively. The intra- and interrun coefficients of variation were 7% ± 4% and 7% ± 3% for UPLC-MS/MS and 5% ± 3% and 4% ± 2% for bioassay, respectively. An excellent correlation between POS plasma concentrations measured by UPLC-MS/MS and bioassay was found (concordance, 0.96). In 26 hemato-oncological patients receiving oral POS, 27/69 (39%) trough plasma concentrations were lower than 0.5 ?g/ml. The UPLC-MS/MS method and sensitive bioassay offer alternative tools for accurate and precise quantification of the plasma concentrations in patients receiving oral posaconazole.

Evaluating sampling strategies and logistic regression methods for modelling complex land cover changes

The role of land cover change as a significant component of global change has become increasingly recognized in recent decades. Large databases measuring land cover change, and the data which can potentially be used to explain the observed changes, are also becoming more commonly available. When developing statistical models to investigate observed changes, it is important to be aware that the chosen sampling strategy and modelling techniques can influence results. We present a comparison of three sampling strategies and two forms of grouped logistic regression models (multinomial and ordinal) in the investigation of patterns of successional change after agricultural land abandonment in Switzerland. Results indicated that both ordinal and nominal transitional change occurs in the landscape and that the use of different sampling regimes and modelling techniques as investigative tools yield different results. Synthesis and applications. Our multimodel inference identified successfully a set of consistently selected indicators of land cover change, which can be used to predict further change, including annual average temperature, the number of already overgrown neighbouring areas of land and distance to historically destructive avalanche sites. This allows for more reliable decision making and planning with respect to landscape management. Although both model approaches gave similar results, ordinal regression yielded more parsimonious models that identified the important predictors of land cover change more efficiently. Thus, this approach is favourable where land cover change pattern can be interpreted as an ordinal process. Otherwise, multinomial logistic regression is a viable alternative.

Visual binding abilities in the initial and advanced stages of schizophrenia

OBJECTIVE: The study tests the hypothesis that intramodal visual binding is disturbed in schizophrenia and should be detectable in all illness stages as a stable trait marker. METHOD: Three groups of patients (rehospitalized chronic schizophrenic, first admitted schizophrenic and schizotypal patients believed to be suffering from a pre-schizophrenic prodrome) and a group of normal control subjects were tested on three tasks targeting visual 'binding' abilities (Muller-Lyer's illusion and two figure detection tasks) in addition to control parameters such as reaction time, visual selective attention, Raven's test and two conventional cortical tasks of spatial working memory (SWM) and a global local test. RESULTS: Chronic patients had a decreased performance on the binding tests. Unexpectedly, the prodromal group exhibited an enhanced Gestalt extraction on these tests compared both to schizophrenic patients and to healthy subjects. Furthermore, chronic schizophrenia was associated with a poor performance on cortical tests of SWM, global local and on Raven. This association appears to be mediated by or linked to the chronicity of the illness. CONCLUSION: The study confirms a variety of neurocognitive deficits in schizophrenia which, however, in this sample seem to be linked to chronicity of illness. However, certain aspects of visual processing concerned with Gestalt extraction deserve attention as potential vulnerability- or prodrome- indicators. The initial hypothesis of the study is rejected.

Energy expenditure by doubly labeled water: validation in humans and proposed calculation.

To further validate the doubly labeled water method for measurement of CO2 production and energy expenditure in humans, we compared it with near-continuous respiratory gas exchange in nine healthy young adult males. Subjects were housed in a respiratory chamber for 4 days. Each received 2H2(18)O at either a low (n = 6) or a moderate (n = 3) isotope dose. Low and moderate doses produced initial 2H enrichments of 5 and 10 X 10(-3) atom percent excess, respectively, and initial 18O enrichments of 2 and 2.5 X 10(-2) atom percent excess, respectively. Total body water was calculated from isotope dilution in saliva collected at 4 and 5 h after the dose. CO2 production was calculated by the two-point method using the isotopic enrichments of urines collected just before each subject entered and left the chamber. Isotope enrichments relative to predose samples were measured by isotope ratio mass spectrometry. At low isotope dose, doubly labeled water overestimated average daily energy expenditure by 8 +/- 9% (SD) (range -7 to 22%). At moderate dose the difference was reduced to +4 +/- 5% (range 0-9%). The isotope elimination curves for 2H and 18O from serial urines collected from one of the subjects showed expected diurnal variations but were otherwise quite smooth. The overestimate may be due to approximations in the corrections for isotope fractionation and isotope dilution. An alternative approach to the corrections is presented that reduces the overestimate to 1%.

The value of selected in vitro and in silico methods to predict acute oral toxicity in a regulatory context: results from the European Project ACuteTox.

ACuteTox is a project within the 6th European Framework Programme which had as one of its goals to develop, optimise and prevalidate a non-animal testing strategy for predicting human acute oral toxicity. In its last 6 months, a challenging exercise was conducted to assess the predictive capacity of the developed testing strategies and final identification of the most promising ones. Thirty-two chemicals were tested blind in the battery of in vitro and in silico methods selected during the first phase of the project. This paper describes the classification approaches studied: single step procedures and two step tiered testing strategies. In summary, four in vitro testing strategies were proposed as best performing in terms of predictive capacity with respect to the European acute oral toxicity classification. In addition, a heuristic testing strategy is suggested that combines the prediction results gained from the neutral red uptake assay performed in 3T3 cells, with information on neurotoxicity alerts identified by the primary rat brain aggregates test method. Octanol-water partition coefficients and in silico prediction of intestinal absorption and blood-brain barrier passage are also considered. This approach allows to reduce the number of chemicals wrongly predicted as not classified (LD50>2000 mg/kg b.w.).

Application of real-time PCR for total airborne bacterial assessment : comparison with epifluorescence microscopy and culture-dependent methods

Traditional culture-dependent methods to quantify and identify airborne microorganisms are limited by factors such as short-duration sampling times and inability to count nonculturableor non-viable bacteria. Consequently, the quantitative assessment of bioaerosols is often underestimated. Use of the real-time quantitative polymerase chain reaction (Q-PCR) to quantify bacteria in environmental samples presents an alternative method, which should overcome this problem. The aim of this study was to evaluate the performance of a real-time Q-PCR assay as a simple and reliable way to quantify the airborne bacterial load within poultry houses and sewage treatment plants, in comparison with epifluorescencemicroscopy and culture-dependent methods. The estimates of bacterial load that we obtained from real-time PCR and epifluorescence methods, are comparable, however, our analysis of sewage treatment plants indicate these methods give values 270-290 fold greater than those obtained by the ''impaction on nutrient agar'' method. The culture-dependent method of air impaction on nutrient agar was also inadequate in poultry houses, as was the impinger-culture method, which gave a bacterial load estimate 32-fold lower than obtained by Q-PCR. Real-time quantitative PCR thus proves to be a reliable, discerning, and simple method that could be used to estimate airborne bacterial load in a broad variety of other environments expected to carry high numbers of airborne bacteria. [Authors]

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.

Analyse critique des résultats de biopsies musculaires et de la méthodologie utilisée dans 889 cas d'affections neuromusculaires [Critical analysis of the results of muscle biopsies and of the methods used in 889 cases of neuromuscular disorders]

The contribution of muscle biopsies to the diagnosis of neuromuscular disorders and the indications of various methods of examination are investigated by analysis of 889 biopsies from patients suffering from myopathic and/or neurogenic disorders. Histo-enzymatic studies performed on frozen material as well as immunohistochemistry and electron microscopy allowed to provide specific diagnoses in all the neurogenic disorders (polyneuropathies and motor neuron diseases), whereas one third of myopathies remained uncertain. Confrontation of neuropathological data with the clinical indications for histological investigations shows that muscle biopsies reveal the diagnosis in 25% of the cases (mainly in congenital and metabolic myopathies) and confirm and/or complete the clinical diagnosis in 50%. In the remaining cases with non specific abnormalities neuropathological investigations may help the clinician by excluding well defined neuromuscular disorders. Analysis of performed studies and results of investigations show the contribution and specificity of each method for the diagnosis. Statistical evaluation of this series indicates that cryostat sectioning for histo- and immunochemical and electron microscopy increases the rate of diagnoses of neuromuscular diseases: full investigation was necessary for the diagnosis in 30% of the cases. The interpretation of the wide range of pathological reactions in muscles requires a close cooperation with the clinician.

Comparison and validation of tissue modelization and statistical classification methods in T1-weighted MR brain images.

This paper presents a validation study on statistical nonsupervised brain tissue classification techniques in magnetic resonance (MR) images. Several image models assuming different hypotheses regarding the intensity distribution model, the spatial model and the number of classes are assessed. The methods are tested on simulated data for which the classification ground truth is known. Different noise and intensity nonuniformities are added to simulate real imaging conditions. No enhancement of the image quality is considered either before or during the classification process. This way, the accuracy of the methods and their robustness against image artifacts are tested. Classification is also performed on real data where a quantitative validation compares the methods' results with an estimated ground truth from manual segmentations by experts. Validity of the various classification methods in the labeling of the image as well as in the tissue volume is estimated with different local and global measures. Results demonstrate that methods relying on both intensity and spatial information are more robust to noise and field inhomogeneities. We also demonstrate that partial volume is not perfectly modeled, even though methods that account for mixture classes outperform methods that only consider pure Gaussian classes. Finally, we show that simulated data results can also be extended to real data.

Concomitant cisplatin and hyperfractionated radiotherapy in locally advanced head and neck cancer: 10-year follow-up of a randomized phase III trial (SAKK 10/94).

Purpose: To compare the long-term outcome of treatment with concomitant cisplatin and hyperfractionated radiotherapy versus treatment with hyperfractionated radiotherapy alone in patients with locally advanced head and neck cancer.Methods and Materials: From July 1994 to July 2000, a total of 224 patients with squamous cell carcinoma of the head and neck were randomized to receive either hyperfractionated radiotherapy alone (median total dose, 74.4 Gy; 1.2 Gy twice daily; 5 days per week) or the same radiotherapy combined with two cycles of cisplatin (20 mg/m(2) for 5 consecutive days during weeks 1 and 5). The primary endpoint was the time to any treatment failure; secondary endpoints were locoregional failure, metastatic failure, overall survival, and late toxicity assessed according to Radiation Therapy Oncology Group criteria.Results: Median follow-up was 9.5 years (range, 0.1-15.4 years). Median time to any treatment failure was not significantly different between treatment arms (hazard ratio [HR], 1.2 [95% confidence interval [CM 0.9-1.7; p = 0.17]). Rates of locoregional failure-free survival (HR, 1.5 [95% CI, 1.1-2.1;p = 0.021), distant metastasis-free survival (HR, 1.6 [95% CI, 1.1-2.5; p = 0.021), and cancer-specific survival (HR, 1.6 [95% CI, 1.0-2.5;p = 0.03]) were significantly improved in the combined-treatment arm, with no difference in major late toxicity between treatment arms. However, overall survival was not significantly different (HR, 1.3 [95% CI, 0.9-1.8; p = 0.11]).Conclusions: After long-term follow-up, combined-treatment with cisplatin and hyperfractionated radiotherapy maintained improved rates of locoregional control, distant metastasis-free survival, and cancer-specific survival compared to that of hyperfractionated radiotherapy alone, with no difference in major late toxicity. (C) 2012 Elsevier Inc.

Carboplatin and weekly paclitaxel doublet chemotherapy compared with monotherapy in elderly patients with advanced non-small-cell lung cancer: IFCT-0501 randomised, phase 3 trial.

BACKGROUND: Platinum-based doublet chemotherapy is recommended to treat advanced non-small-cell lung cancer (NSCLC) in fit, non-elderly adults, but monotherapy is recommended for patients older than 70 years. We compared a carboplatin and paclitaxel doublet chemotherapy regimen with monotherapy in elderly patients with advanced NSCLC. METHODS: In this multicentre, open-label, phase 3, randomised trial we recruited patients aged 70-89 years with locally advanced or metastatic NSCLC and WHO performance status scores of 0-2. Patients received either four cycles (3 weeks on treatment, 1 week off treatment) of carboplatin (on day 1) plus paclitaxel (on days 1, 8, and 15) or five cycles (2 weeks on treatment, 1 week off treatment) of vinorelbine or gemcitabine monotherapy. Randomisation was done centrally with the minimisation method. The primary endpoint was overall survival, and analysis was done by intention to treat. This trial is registered, number NCT00298415. FINDINGS: 451 patients were enrolled. 226 were randomly assigned monotherapy and 225 doublet chemotherapy. Median age was 77 years and median follow-up was 30.3 months (range 8.6-45.2). Median overall survival was 10.3 months for doublet chemotherapy and 6.2 months for monotherapy (hazard ratio 0.64, 95% CI 0.52-0.78; p<0.0001); 1-year survival was 44.5% (95% CI 37.9-50.9) and 25.4% (19.9-31.3), respectively. Toxic effects were more frequent in the doublet chemotherapy group than in the monotherapy group (most frequent, decreased neutrophil count (108 [48.4%] vs 28 [12.4%]; asthenia 23 [10.3%] vs 13 [5.8%]). INTERPRETATION: Despite increased toxic effects, platinum-based doublet chemotherapy was associated with survival benefits compared with vinorelbine or gemcitabine monotherapy in elderly patients with NSCLC. We feel that the current treatment paradigm for these patients should be reconsidered. FUNDING: Intergroupe Francophone de Cancérologie Thoracique, Institut National du Cancer.