Positron emission tomography (PET) is a powerful metabolic imaging modality, which is designed to detect two anti-parallel 511 keV photons origniating from a positron-electron annihilation. However, it is possible that one or both of the annihilation photons undergo a Compton scattering in the object. This is more serious for a scanner operated in 3D mode or with large patients, where the scatter fraction can be as high as 40-60%. When one or both photons are scattered, the line of response (LOR) defined by connecting the two relevant detectors no longer passes through the annihilation position. Thus, scattered coincidences degrade image contrast and compromise quantitative accuracy. Various scatter correction methods have been proposed but most of them are based on estimating and subtracting the scatter from the measured data or incorporating it into an iterative reconstruction algorithm. By accurately measuring the scattered photon energy and taking advantage of the kinematics of Compton scattering, two circular arcs (TCA) in 2D can be identified, which describe the locus of all the possible scattering positions and encompass the point of annihilation. In the limiting case where the scattering angle approaches zero, the TCA approach the LOR for true coincidences. Based on this knowledge, a Generalized Scatter (GS) reconstruction algorithm has been developed in this thesis, which can use both true and scattered coincidences to extract the activity distribution in a consistent way. The annihilation position within the TCA can be further confined by adding a patient outline as a constraint into the GS algorithm. An attenuation correction method for the scattered coincidences was also developed in order to remove the imaging artifacts. A geometrical model that characterizes the different probabilities of the annihilation positions within the TCA was also proposed. This can speed up image convergence and improve reconstructed image quality. Finally, the GS algorithm has been adapted to deal with non-ideal energy resolutions. In summary, an algorithm that implicitly incorporates scattered coincidences into the image reconstruction has been developed. Our results demonstrate that this eliminates the need for scatter correction and can improve system sensitivity and image quality. / February 2016

Orientador: João Pedro Albino / Resumo: Os Sistemas de Informações Geográficas – SIG – têm sido muito utilizados pelos órgãos de saúde para análise e tomada de decisão nas ações de combate a diversas doenças e têm alcançado resultados dentro dos parâmetros de controle dos vetores, porém essas análises não permitem uma fácil verificação de tendências e evolução das mesmas, ocasionando demora na tomada de decisão e, comumente, um superdimensionamento dos recursos utilizados. Com o objetivo de criar uma visualização mais eficaz, foi desenvolvido um método de análise e representação de dados de endemias que permitiu o desenvolvido de mapas interativos capazes de representar a densidade, a concentração e a dispersão dos casos (mapa de calor - heatmap), utilizando o Sistema R e Python. Ao aplicar este método baseado nos dados de casos de leishmaniose humana, fornecidos pela Secretaria da Saúde na cidade de Marília, interior de São Paulo, no período de 2011 a 2019, foi possível identificar com mais precisão em quais locais estão ocorrendo a transmissão animal-humana. A interação na linha do tempo do mapa permitiu verificar o comportamento dessa contaminação, sendo possível prever por quais regiões os casos devem se alastrar e quais os principais pontos de bloqueio do avanço da doença devem ser definidos. / Abstract: Geographic Information Systems - GIS - have been widely used by health agencies for analysis and decision making in actions to combat various diseases and have achieved results within the parameters of vector control, however these analyzes do not allow an easy verification of trends and their evolution, causing delay in decision making and, commonly, an oversizing of the resources used. In order to create a more effective visualization, a method of analysis and representation of endemic data was developed that allowed the creation of interactive maps capable of representing the density, concentration and dispersion of the cases (density map - heatmap), using the R and Python System. By applying this method based on data from cases of human leishmaniasis, provided by the Secretariat of Health in the city of Marília, in the interior of São Paulo, from 2011 to 2019, it was possible to identify more precisely in which locations animal-human transmission is occurring. The interaction in the timeline of the map allowed to verify the behavior of this contamination, making it possible to predict which regions the cases should spread to and which are the main points of blocking the progress of the disease. / Mestre

Proteins are involved in almost all functions in a living cell, and functions of proteins are realized by their tertiary structures. Protein three-dimensional structures can be solved by multiple experimental methods, but computational approaches serve as an important complement to experimental methods for comparing and analyzing protein structures. Protein structure comparison allows the transfer of knowledge about known proteins to a novel protein and plays an important role in function prediction. Obtaining a global perspective of the variety and distribution of protein structures also lays a foundation for our understanding of the building principle of protein structures. This dissertation introduces our computational method to compare protein 3D structures and presents a novel mapping of protein shapes that represents the variety and the similarities of 3D shapes of proteins and their assemblies. The methods developed in this work can be applied to obtain new biological insights into protein atomic structures and electron density maps.

In spatial analysis, many operations are performed considering the neighbouring locations of a feature. The standard definition of a neighbourhood is an area confined by geometrical length and direction with respect to its focus. When allocating a location for a service, the population distribution is often considered. Standard GIS software includes tools for computations with uniform neighbourhoods, usually equal sized circles. These tools can be used for distribution analysis. Many geographic studies used as basis for city planning decisions use distance as an evaluator. It is a frequent occurrence that the actual distance is approximated using factored straight-line distance. For great distances and large datasets, this is a sufficient means of evaluation, whilst for smaller distances for specific locations, it poses major drawbacks. For distribution analysis in a network space, the neighbourhood would need to be derived from the local set of network features, creating a unique neighbourhood for each location. The neighbourhood can then be used to overlay other datasets to perform analysis of features within the network space. This report describes the application of network distance based neighbourhoods to design a tool, Network Location Analysis, for calculating focal statistics for use as a city planning decision support. The tool has been implemented as a workflow of ArcGIS tools scripted as a Python toolbox. The input required by the tool is a population point layer and a vector network dataset. The output is a grid of points with population statistics as attributes and corresponding neighbourhoods generalized as polygons. The tool has been tested by comparing it to standard focal operations implemented in ArcGIS and by applying it to the dataset used when conducting a study on the location of a new metro station using conventional ArcGIS tools. The results have been analysed and visualized and compared to data used in the study. The result concludes that Network Location Analysis surpasses conventional ArcGIS tools when conducting analysis on features in a network. It derives an accurate set of sum and proximity statistics for all locations within the processing extent, enabling analysis on the population distribution throughout the area and for specific points. The output is intuitive, manageable and can be visualized as raster or by displaying the neighbourhoods as polygons and can be used to evaluate population distribution and network connectivity. The drawbacks of the tool are its lack of robustness, its rigidity to input and the inefficient implementation causing execution time to be unpractical.

In this thesis, we study the problem of protein side chain assignment (SCA) given multiple sources of experimental and modeling data. In particular, the mechanism of X-ray crystallography (X-ray) is re-examined using Fourier analysis, and a novel probabilistic model of X-ray is proposed for SCA's decision making. The relationship between the measurements in X-ray and the desired structure is reformulated in terms of Discrete Fourier Transform (DFT). The decision making is performed by developing a new resolution-dependent electron density map (EDM) model and applying Maximum Likelihood (ML) estimation, which simply reduces to the Least Squares (LS) solution. Calculation of the condence probability associated with this decision making is also given. One possible extension of this novel model is the real-space refinement when the continuous conformational space is used. Furthermore, we present a data fusion scheme combining multi-sources of data to solve SCA problem. The merit of our framework is the capability of exploiting multi-sources of information to make decisions in a probabilistic perspective based on Bayesian inference. Although our approach aims at SCA problem, it can be easily transplanted to solving for the entire protein structure.

Tato práce se zaměřuje na problém počítání vozidel v statickém obraze bez znalosti geometrických vlastností scény. V rámci řešení bylo implementováno a natrénováno 5 architektur konvolučních neuronových sítí. Také byl pořízen rozsáhlý dataset s 19 310 snímky pořízených z 12pohledů a zachycujících 7 různých scén. Použité konvoluční sítě mapují vstupní vzorek na mapu hustoty vozidel, ze které lze získat jejich počet a lokalizaci v kontextu vstupního snímku. Hlavním přínosem této práce je porovnání a aplikace dosavadních nejlepších řešení pro počítání objektů v obraze. Většina z těchto architektur byla navržena pro počítání lidí v obraze, proto musely být uzpůsobeny pro potřeby počítání vozidel v statickém obraze. Natrénované modely jsou vyhodnoceny GAME metrikou na TRANCOS datasetu a na velkém spojeném datasetu. Dosažené výsledky všech modelů jsou následně popsány a porovnány.