Implementación de una Support Vector Machine en RVC – CAL para imágenes hiperespectrales

El análisis de imágenes hiperespectrales permite obtener información con una gran resolución espectral: cientos de bandas repartidas desde el espectro infrarrojo hasta el ultravioleta. El uso de dichas imágenes está teniendo un gran impacto en el campo de la medicina y, en concreto, destaca su utilización en la detección de distintos tipos de cáncer. Dentro de este campo, uno de los principales problemas que existen actualmente es el análisis de dichas imágenes en tiempo real ya que, debido al gran volumen de datos que componen estas imágenes, la capacidad de cómputo requerida es muy elevada. Una de las principales líneas de investigación acerca de la reducción de dicho tiempo de procesado se basa en la idea de repartir su análisis en diversos núcleos trabajando en paralelo. En relación a esta línea de investigación, en el presente trabajo se desarrolla una librería para el lenguaje RVC – CAL – lenguaje que está especialmente pensado para aplicaciones multimedia y que permite realizar la paralelización de una manera intuitiva – donde se recogen las funciones necesarias para implementar el clasificador conocido como Support Vector Machine – SVM. Cabe mencionar que este trabajo complementa el realizado en [1] y [2] donde se desarrollaron las funciones necesarias para implementar una cadena de procesado que utiliza el método unmixing para procesar la imagen hiperespectral. En concreto, este trabajo se encuentra dividido en varias partes. La primera de ellas expone razonadamente los motivos que han llevado a comenzar este Trabajo de Investigación y los objetivos que se pretenden conseguir con él. Tras esto, se hace un amplio estudio del estado del arte actual y, en él, se explican tanto las imágenes hiperespectrales como sus métodos de procesado y, en concreto, se detallará el método que utiliza el clasificador SVM. Una vez expuesta la base teórica, nos centraremos en la explicación del método seguido para convertir una versión en Matlab del clasificador SVM optimizado para analizar imágenes hiperespectrales; un punto importante en este apartado es que se desarrolla la versión secuencial del algoritmo y se asientan las bases para una futura paralelización del clasificador. Tras explicar el método utilizado, se exponen los resultados obtenidos primero comparando ambas versiones y, posteriormente, analizando por etapas la versión adaptada al lenguaje RVC – CAL. Por último, se aportan una serie de conclusiones obtenidas tras analizar las dos versiones del clasificador SVM en cuanto a bondad de resultados y tiempos de procesado y se proponen una serie de posibles líneas de actuación futuras relacionadas con dichos resultados. ABSTRACT. Hyperspectral imaging allows us to collect high resolution spectral information: hundred of bands covering from infrared to ultraviolet spectrum. These images have had strong repercussions in the medical field; in particular, we must highlight its use in cancer detection. In this field, the main problem we have to deal with is the real time analysis, because these images have a great data volume and they require a high computational power. One of the main research lines that deals with this problem is related with the analysis of these images using several cores working at the same time. According to this investigation line, this document describes the development of a RVC – CAL library – this language has been widely used for working with multimedia applications and allows an optimized system parallelization –, which joins all the functions needed to implement the Support Vector Machine – SVM - classifier. This research complements the research conducted in [1] and [2] where the necessary functions to implement the unmixing method to analyze hyperspectral images were developed. The document is divided in several chapters. The first of them introduces the motivation of the Master Thesis and the main objectives to achieve. After that, we study the state of the art of some technologies related with this work, like hyperspectral images, their processing methods and, concretely, the SVM classifier. Once we have exposed the theoretical bases, we will explain the followed methodology to translate a Matlab version of the SVM classifier optimized to process an hyperspectral image to RVC – CAL language; one of the most important issues in this chapter is that a sequential implementation is developed and the bases of a future parallelization of the SVM classifier are set. At this point, we will expose the results obtained in the comparative between versions and then, the results of the different steps that compose the SVM in its RVC – CAL version. Finally, we will extract some conclusions related with algorithm behavior and time processing. In the same way, we propose some future research lines according to the results obtained in this document.

Nuevos desarrollos en la medida de descargas parciales y técnicas avanzadas de análisis para el diagnóstico del estado de los aislamientos en sistemas eléctricos de alta tensión

En esta tesis se desarrolla una investigación sobre las técnicas de medida de descargas parciales (DP) que se aplican en medidas on-line, en condiciones normales de operación de las instalaciones eléctricas de alta tensión (AT). También se realiza un estudio de técnicas avanzadas de procesado y análisis de las señales medidas, que permiten realizar diagnósticos precisos del estado de los aislamientos eléctricos de AT. Uno de los objetivos fundamentales de la tesis ha sido disponer de un procedimiento eficaz de medida y procesado de las señales de DP, para la realización de medidas on-line tanto de forma itinerante, como mediante monitorización temporal o permanente. La implementación del nuevo procedimiento de medida permite obtener resultados satisfactorios en la detección, identificación y localización de defectos de aislamiento. Se ha dedicado especial interés al desarrollo de un método de clasificación de señales, que permite separar pulsos de ruido y diferentes fuentes de DP, presentes de forma simultánea en las instalaciones de AT. El estudio de la clasificación de señales se ha completado con la aplicación de un método para la detección de manera asistida, de los diferentes grupos de pulsos de ruido y de DP. La aplicación de este método de detección de grupos de pulsos, facilita la labor de los técnicos especialistas a la hora de diagnosticar el estado de los elementos aislantes. Al efecto de verificar de forma práctica las aportaciones de la tesis, se han realizado medidas de DP tanto en laboratorio como en campo. Las medidas experimentales en laboratorio se han efectuado en el Laboratorio de Alta Tensión de la Escuela Técnica Superior de Ingeniería y Diseño Industrial (LAT-UPM), de la Universidad Politécnica de Madrid. Por otro lado, las medidas experimentales en campo se han llevado a cabo en instalaciones de AT propiedad de compañías de transporte y distribución de energía eléctrica. La realización de ensayos de DP en estas instalaciones ha sido posible, gracias a los proyectos de investigación llevados a cabo por el grupo de trabajo del LAT-UPM, con diferentes empresas del sector durante los diez últimos años. ABSTRACT This thesis develops techniques for measuring partial discharges (PD) that are applied in on-line measurements, under normal operating conditions of the high voltage (HV) electrical installations. In addition there are studied advanced techniques for the processing and analysis of the measured signals, that permit precise diagnostics of the state of HV electrical insulation systems. One of the fundamental objectives of the thesis is to make available an effective procedure for measuring and processing PD signals, for making on-line measurements, either in an itinerant way or in temporary or permanent monitoring. The implementation of the new measurement procedure yields satisfactory results in the detection, identification and localization of insulation defects. Special attention has been devoted to the development of a method for classifying signals, that separates noise pulses and various PD sources present simultaneously in the HV installations. The study of the classification of signals has been completed by the application of a method for detecting, in a user assisted manner, the different groups of noise pulses and of PD. The application of this method for detecting groups of pulses facilitates the work of the specialist technicians to diagnose the condition of the insulation elements. To demonstrate the practical value of the thesis, PD measurements were made in laboratory as well as in field installations. The experimental measurements in laboratory were made in the High Voltage Laboratory (LAT-UPM) of the High Technical School of Engineering and Industrial Design, of the Polytechnic University of Madrid. Field measurements were realized in the HV installations of companies providing electrical energy transport and distribution. The realization of PD tests in these facilities was possible thanks to the research projects carried out by the working group of the LAT-UPM during the last ten years, with different companies operating in the sector.

Higher temporal variability of forest breeding bird communities in fragmented landscapes

Understanding the relationship between animal community dynamics and landscape structure has become a priority for biodiversity conservation. In particular, predicting the effects of habitat destruction that confine species to networks of small patches is an important prerequisite to conservation plan development. Theoretical models that predict the occurrence of species in fragmented landscapes, and relationships between stability and diversity do exist. However, reliable empirical investigations of the dynamics of biodiversity have been prevented by differences in species detection probabilities among landscapes. Using long-term data sampled at a large spatial scale in conjunction with a capture-recapture approach, we developed estimates of parameters of community changes over a 22-year period for forest breeding birds in selected areas of the eastern United States. We show that forest fragmentation was associated not only with a reduced number of forest bird species, but also with increased temporal variability in the number of species. This higher temporal variability was associated with higher local extinction and turnover rates. These results have major conservation implications. Moreover, the approach used provides a practical tool for the study of the dynamics of biodiversity.

Functional and physiological consequences of genetic variation at phosphoglucose isomerase: Heat shock protein expression is related to enzyme genotype in a montane beetle

Allele frequency variation at the phosphoglucose isomerase (PGI) locus in Californian populations of the beetle Chrysomela aeneicollis suggests that PGI may be undergoing natural selection. We quantified (i) apparent Michaelis-Menten constant (Km) of fructose 6-phosphate at different temperatures and (ii) thermal stability for three common PGI genotypes (1–1, 1–4, and 4–4). We also measured air temperature (Ta) and beetle body temperature (Tb) in three montane drainages in the Sierra Nevada, California. Finally, we measured 70-kDa heat shock protein (Hsp70) expression in field-collected and laboratory-acclimated beetles. We found that PGI allele 1 predominated in the northernmost drainage, Rock Creek (RC), which was also significantly cooler than the southernmost drainage, Big Pine Creek (BPC), where PGI allele 4 predominated. Allele frequencies and air temperatures were intermediate in the middle drainage, Bishop Creek (BC). Differences among genotypes in Km (1–1 > 1–4 > 4–4) and thermal stability (4–4 > 1–4 > 1–1) followed a pattern consistent with temperature adaptation. In nature, Tb was closely related to Ta. Hsp70 expression in adult beetles decreased with elevation and differed among drainages (BPC > BC > RC). After laboratory acclimation (8 days, 20°C day, 4°C night) and heat shock (4 h, 28–36°C), Hsp70 expression was greater for RC than BPC beetles. In RC, field-collected beetles homozygous for PGI 1–1 had higher Hsp70 levels than heterozygotes or a 4–4 homozygote. These results reveal functional and physiological differences among PGI genotypes, which suggest that montane populations of this beetle are locally adapted to temperature.

Ambient UV-B radiation causes deformities in amphibian embryos

There has been a great deal of recent attention on the suspected increase in amphibian deformities. However, most reports of amphibian deformities have been anecdotal, and no experiments in the field under natural conditions have been performed to investigate this phenomenon. Under laboratory conditions, a variety of agents can induce deformities in amphibians. We investigated one of these agents, UV-B radiation, in field experiments, as a cause for amphibian deformities. We monitored hatching success and development in long-toed salamanders under UV-B shields and in regimes that allowed UV-B radiation. Embryos under UV-B shields had a significantly higher hatching rate and fewer deformities, and developed more quickly than those exposed to UV-B. Deformities may contribute directly to embryo mortality, and they may affect an individual’s subsequent survival after hatching.

Cnr interferes with dimerization of the replication protein α in phage-plasmid P4

DNA replication of phage-plasmid P4 in its host Escherichia coli depends on its replication protein α. In the plasmid state, P4 copy number is controlled by the regulator protein Cnr (copy number regulation). Mutations in α (αcr) that prevent regulation by Cnr cause P4 over-replication and cell death. Using the two-hybrid system in Saccharomyces cerevisiae and a system based on λ immunity in E.coli for in vivo detection of protein–protein interactions, we found that: (i) α protein interacts with Cnr, whereas αcr proteins do not; (ii) both α–α and αcr–αcr interactions occur and the interaction domain is located within the C-terminal of α; (iii) Cnr–Cnr interaction also occurs. Using an in vivo competition assay, we found that Cnr interferes with both α–α and αcr–αcr dimerization. Our data suggest that Cnr and α interact in at least two ways, which may have different functional roles in P4 replication control.

Methylation is not the main force repressing the retrotransposon MAGGY in Magnaporthe grisea

We have introduced the LTR-retrotransposon MAGGY into a naive genome of Magnaporthe grisea and estimated the copy number of MAGGY in a cell by serial isolation of fungal protoplasts at certain time intervals. The number of MAGGY elements rapidly increased for a short period following introduction. However, it did not increase geometrically and reached equilibrium at 20–30 copies per genome, indicating that MAGGY was repressed or silenced during proliferation. De novo methylation of MAGGY occurred immediately following invasion into the genome but the degree of methylation was constant and did not correlate with the repression of MAGGY. 5-Azacytidine treatment demethylated and transcriptionally activated the MAGGY element in regenerants but did not affect transpositional frequency, suggesting that post-transcriptional suppression, not methylation, is the main force that represses MAGGY proliferation in M.grisea. Support for this conclusion was also obtained by examining the methylation status of MAGGY sequences in field isolates of M.grisea with active or inactive MAGGY elements. Methylation of the MAGGY sequences was detected in some isolates but not in others. However, the methylation status did not correlate with the copy numbers and activity of the elements.

Visual constraints in foraging bumblebees: Flower size and color affect search time and flight behavior

In optimal foraging theory, search time is a key variable defining the value of a prey type. But the sensory-perceptual processes that constrain the search for food have rarely been considered. Here we evaluate the flight behavior of bumblebees (Bombus terrestris) searching for artificial flowers of various sizes and colors. When flowers were large, search times correlated well with the color contrast of the targets with their green foliage-type background, as predicted by a model of color opponent coding using inputs from the bees' UV, blue, and green receptors. Targets that made poor color contrast with their backdrop, such as white, UV-reflecting ones, or red flowers, took longest to detect, even though brightness contrast with the background was pronounced. When searching for small targets, bees changed their strategy in several ways. They flew significantly slower and closer to the ground, so increasing the minimum detectable area subtended by an object on the ground. In addition, they used a different neuronal channel for flower detection. Instead of color contrast, they used only the green receptor signal for detection. We relate these findings to temporal and spatial limitations of different neuronal channels involved in stimulus detection and recognition. Thus, foraging speed may not be limited only by factors such as prey density, flight energetics, and scramble competition. Our results show that understanding the behavioral ecology of foraging can substantially gain from knowledge about mechanisms of visual information processing.

Induction of long-term potentiation is associated with major ultrastructural changes of activated synapses.

Long-term potentiation (LTP), an increase in synaptic efficacy believed to underlie learning and memory mechanisms, has been proposed to involve structural modifications of synapses. Precise identification of the morphological changes associated with LTP has however been hindered by the difficulty in distinguishing potentiated or activated from nonstimulated synapses. Here we used a cytochemical method that allowed detection in CA1 hippocampus at the electron microscopy level of a stimulation-specific, D-AP5-sensitive accumulation of calcium in postsynaptic spines and presynaptic terminals following application of high-frequency trains. Morphometric analyses carried out 30-40 min after LTP induction revealed dramatic ultrastructural differences between labeled and nonlabeled synapses. The majority of labeled synapses (60%) exhibited perforated postsynaptic densities, whereas this proportion was only 20% in nonlabeled synaptic contacts. Labeled synaptic profiles were also characterized by a larger apposition zone between pre- and postsynaptic structures, longer postsynaptic densities, and enlarged spine profiles. These results add strong support to the idea that ultrastructural modifications and specifically an increase in perforated synapses are associated with LTP induction in field CA1 of hippocampus and they suggest that a majority of activated contacts may exhibit such changes.

Sistema para monitoramento e análise de paisagens acústicas submarinas.

O Monitoramento Acústico Passivo (PAM) submarino refere-se ao uso de sistemas de escuta e gravação subaquática, com o intuito de detectar, monitorar e identificar fontes sonoras através das ondas de pressão que elas produzem. Se diz que é passivo já que tais sistemas unicamente ouvem, sem perturbam o meio ambiente acústico existente, diferentemente de ativos, como os sonares. O PAM submarino tem diversas áreas de aplicação, como em sistemas de vigilância militar, seguridade portuária, monitoramento ambiental, desenvolvimento de índices de densidade populacional de espécies, identificação de espécies, etc. Tecnologia nacional nesta área é praticamente inexistente apesar da sua importância. Neste contexto, o presente trabalho visa contribuir com o desenvolvimento de tecnologia nacional no tema através da concepção, construção e operação de equipamento autônomo de PAM e de métodos de processamento de sinais para detecção automatizada de eventos acústicos submarinos. Foi desenvolvido um equipamento, nomeado OceanPod, que possui características como baixo custo de fabrica¸c~ao, flexibilidade e facilidade de configuração e uso, voltado para a pesquisa científica, industrial e para controle ambiental. Vários protótipos desse equipamento foram construídos e utilizados em missões no mar. Essas jornadas de monitoramento permitiram iniciar a criação de um banco de dados acústico, o qual permitiu fornecer a matéria prima para o teste de detectores de eventos acústicos automatizados e em tempo real. Adicionalmente também é proposto um novo método de detecção-identificação de eventos acústicos, baseado em análise estatística da representação tempo-frequência dos sinais acústicos. Este novo método foi testado na detecção de cetáceos, presentes no banco de dados gerado pelas missões de monitoramento.

Time-motion analysis in women’s team handball: importance of aerobic performance

Women’s handball is a sport, which has seen an accelerated development over the last decade. Data on movement patterns in combination with physiological demands are nearly nonexistent in the literature. The aim of this study was twofold: first, to analyze the horizontal movement pattern, including the sprint acceleration profiles, of individual female elite handball players and the corresponding heart rates (HRs) during a match and secondly to determine underlying correlations with individual aerobic performance. Players from one German First League team (n = 11) and the Norwegian National Team (n = 14) were studied during one match using the Sagit system for movement analysis and Polar HR monitoring for analysis of physiological demands. Mean HR during the match was 86 % of maximum HR (HRmax). With the exception of the goalkeepers (GKs, 78 % of HRmax), no position-specific differences could be detected. Total distance covered during the match was 4614 m (2066 m in GKs and 5251 m in field players (FPs)). Total distance consisted of 9.2 % sprinting, 26.7 % fast running, 28.8 % slow running, and 35.5 % walking. Mean velocity varied between 1.9 km/h (0.52 m/s) (GKs) and 4.2 km/h (1.17 m/s) (FPs, no position effect). Field players with a higher level of maximum oxygen uptake (V̇O2max) executed run activities with a higher velocity but comparable percentage of HRmax as compared to players with lower aerobic performance, independent of FP position. Acceleration profile depended on aerobic performance and the field player’s position. In conclusion, a high V̇O2max appears to be important in top-level international women’s handball. Sprint and endurance training should be conducted according to the specific demands of the player’s position.

Screen-printed electrode-based electrochemical detector coupled with in-situ ionic-liquid-assisted dispersive liquid–liquid microextraction for determination of 2,4,6-trinitrotoluene

A novel method is reported, whereby screen-printed electrodes (SPELs) are combined with dispersive liquid–liquid microextraction. In-situ ionic liquid (IL) formation was used as an extractant phase in the microextraction technique and proved to be a simple, fast and inexpensive analytical method. This approach uses miniaturized systems both in sample preparation and in the detection stage, helping to develop environmentally friendly analytical methods and portable devices to enable rapid and onsite measurement. The microextraction method is based on a simple metathesis reaction, in which a water-immiscible IL (1-hexyl-3-methylimidazolium bis[(trifluoromethyl)sulfonyl]imide, [Hmim][NTf2]) is formed from a water-miscible IL (1-hexyl-3-methylimidazolium chloride, [Hmim][Cl]) and an ion-exchange reagent (lithium bis[(trifluoromethyl)sulfonyl]imide, LiNTf2) in sample solutions. The explosive 2,4,6-trinitrotoluene (TNT) was used as a model analyte to develop the method. The electrochemical behavior of TNT in [Hmim][NTf2] has been studied in SPELs. The extraction method was first optimized by use of a two-step multivariate optimization strategy, using Plackett–Burman and central composite designs. The method was then evaluated under optimum conditions and a good level of linearity was obtained, with a correlation coefficient of 0.9990. Limits of detection and quantification were 7 μg L−1 and 9 μg L−1, respectively. The repeatability of the proposed method was evaluated at two different spiking levels (20 and 50 μg L−1), and coefficients of variation of 7 % and 5 % (n = 5) were obtained. Tap water and industrial wastewater were selected as real-world water samples to assess the applicability of the method.

New Approach for Correction of Error Associated With Keratometric Estimation of Corneal Power in Keratoconus

The aim of this study was to obtain the exact value of the keratometric index (nkexact) and to clinically validate a variable keratometric index (nkadj) that minimizes this error. Methods: The nkexact value was determined by obtaining differences (DPc) between keratometric corneal power (Pk) and Gaussian corneal power (PGauss c ) equal to 0. The nkexact was defined as the value associated with an equivalent difference in the magnitude of DPc for extreme values of posterior corneal radius (r2c) for each anterior corneal radius value (r1c). This nkadj was considered for the calculation of the adjusted corneal power (Pkadj). Values of r1c ∈ (4.2, 8.5) mm and r2c ∈ (3.1, 8.2) mm were considered. Differences of True Net Power with PGauss c , Pkadj, and Pk(1.3375) were calculated in a clinical sample of 44 eyes with keratoconus. Results: nkexact ranged from 1.3153 to 1.3396 and nkadj from 1.3190 to 1.3339 depending on the eye model analyzed. All the nkadj values adjusted perfectly to 8 linear algorithms. Differences between Pkadj and PGauss c did not exceed 60.7 D (Diopter). Clinically, nk = 1.3375 was not valid in any case. Pkadj and True Net Power and Pk(1.3375) and Pkadj were statistically different (P , 0.01), whereas no differences were found between PGauss c and Pkadj (P . 0.01). Conclusions: The use of a single value of nk for the calculation of the total corneal power in keratoconus has been shown to be imprecise, leading to inaccuracies in the detection and classification of this corneal condition. Furthermore, our study shows the relevance of corneal thickness in corneal power calculations in keratoconus.

Analyzation of marine snow and fecal pellets collected in sediment trap CBi-2 off Cape Blanc, Mauritania

We analyzed size-specific dry mass, sinking velocity, and apparent diffusivity in field-sampled marine snow, laboratory-made aggregates formed by diatoms or coccolithophorids, and small and large zooplankton fecal pellets with naturally varying content of ballast materials. Apparent diffusivity was measured directly inside aggregates and large (millimeter-long) fecal pellets using microsensors. Large fecal pellets, collected in the coastal upwelling off Cape Blanc, Mauritania, showed the highest volume-specific dry mass and sinking velocities because of a high content of opal, carbonate, and lithogenic material (mostly Saharan dust), which together comprised ~80% of the dry mass. The average solid matter density within these large fecal pellets was 1.7 g cm**-3, whereas their excess density was 0.25 ± 0.07 g cm**-3. Volume-specific dry mass of all sources of aggregates and fecal pellets ranged from 3.8 to 960 µg mm**-3, and average sinking velocities varied between 51 and 732 m d**-1. Porosity was >0.43 and >0.96 within fecal pellets and phytoplankton-derived aggregates, respectively. Averaged values of apparent diffusivity of gases within large fecal pellets and aggregates were 0.74 and 0.95 times that of the free diffusion coefficient in sea water, respectively. Ballast increases sinking velocity and, thus, also potential O2 fluxes to sedimenting aggregates and fecal pellets. Hence, ballast minerals limit the residence time of aggregates in the water column by increasing sinking velocity, but apparent diffusivity and potential oxygen supply within aggregates are high, whereby a large fraction of labile organic carbon can be respired during sedimentation.

Limitations of the widely used GAM42a and BET42a probes targeting bacteria in the Gammaproteobacteria radiation

The 23S rRNA-targeted probes GAM42a and BET42a provided equivocal results with the uncultured gammaproteobacterium 'Candidatus Competibacter phosphatis' where some cells bound GAM42a and other cells bound BET42a in fluorescence in situ hybridization (FISH) experiments. Probes GAM42a and BET42a span positions 1027-1043 in the 23S rRNAand differ from each other by one nucleotide at position 1033. Clone libraries were prepared from PCR products spanning the 16S rRNA genes, intergenic spacer region and 23S rRNA genes from two mixed cultures enriched in 'Candidatus C. phosphatis'. With individual clone inserts, the 16S rDNA portion was used to confirm the source organism as 'Candidatus C. phosphatis' and the 23S rDNA portion was used to determine the sequence of the GAM42a/BET42a probe target region. Of the 19 clones sequenced, 8 had the GAM42a probe target (T at position 1033) and 11 had G at position 1033, the only mismatch with GAM42a. However, none of the clones had the BET42a probe target (A at 1033). Non-canonical base-pairing between the 23S rRNA of 'Candidatus C. phosphatis' with G at position 1033 and GAM42a (G-A) or BET42a (G-T) is likely to explain the probing anomalies. A probe (GAM42_C1033) was optimized for use in FISH, targeting cells with G at position 1033, and was found to highlight not only some 'Candidatus C. phosphatis' cells, but also other bacteria. This demonstrates that there are bacteria in addition to 'Candidatus C. phosphatis' with the GAM42_C1033 probe target and not the BET42a or GAM42a probe target.