A documentação clínica requer a representação de situações complexas como pareceres clínicos, imagens e resultados de exames, planos de tratamento, dentre outras. Entre os profissionais da área de saúde, a linguagem natural é o meio principal de documentação. Neste tipo de linguagem, caracterizada por uma elevada flexibilidade sintática e léxica, é comum a prevalência de ambigüidades em sentenças e termos. O objetivo do presente trabalho consiste em mapear informações codificadas em narrativas clínicas para uma ontologia de domínio (SNOMED CT). Para sua consecução, aplicaram-se ferramentas processamento de linguagem natural (PLN), assim como adotaram-se heurísticas para o mapeamento de textos para ontologias. Para o desenvolvimento da pesquisa, uma amostra de sumários de alta foi obtida junto ao Hospital das Clínicas de Porto Alegre, RS, Brasil. Parte dos sumários foi manualmente anotada, com a aplicação da estratégia de Active Learning, visando a preparação de um corpus para o treinamento de ferramentas de PLN. Paralelamente, foram desenvolvidos algoritmos para o pré-processamento dos textos ‘sujos’ (com grande quantidade de erros, acrônimos, abreviações, etc). Com a identificação das frases nominais, resultado do processamento das ferramentas de PLN, diversas heurísticas (identificação de acrônimos, correção ortográfica, supressão de valores numéricos e distância conceitual) para o mapeamento para a SNOMED CT foram aplicadas. A versão atual da SNOMED CT não está disponível em português, demandando o uso de ferramentas para processamento multi-lingual. Para tanto, o pesquisa atual é parte da iniciativa do projeto MorphoSaurus, por meio do qual desenvolve-se e disponibiliza-se um thesaurus multi-língue (português, alemão, inglês, espanhol, sueco, francês), bem como componentes de software que permitem o processamento inter-lingual. Para realização da pesquisa, 80% da base de sumários foi analisada e manualmente anotada, resultando em um corpus de domínio (textos médicos e em português) que permitiu a especialização do software OpenNLP (baseado no modelo estatístico para o PLN e selecionado após a avaliação de outras soluções disponíveis). A precisão do etiquetador atingiu 93.67%. O thesaurus multi-língue do MorphoSaurus foi estendido, reestruturado e avaliado (automaticamente com a comparação por meio de textos comparáveis – ‘traduções de um mesmo texto para diferentes idiomas’) e sofreu intervenções objetivando a correção de imperfeições existentes, resultando na melhoria da cobertura lingüística, no caso do português, em 2%; e 50% para o caso do espanhol, medidas obtidas por meio do levantamento das curvas de precisão e revocação para a base do OHSUMED. Por fim, a codificação de informações de narrativas clínicas para uma ontologia de domínio é uma área de elevado interesse científico e clínico, visto que grande parte dos dados produzidos quando do atendimento médico é armazenado em texto livre e não em campos estruturados. Para o alcance deste fim, adotou-se a SNOMED CT. A viabilidade da metodologia de mapeamento foi demonstrada com a avaliação dos resultados do mapeamento automático contra um padrão ouro, manualmente desenvolvido, indicando precisão de 83,9%. / Clinical documentation requires the representation of fine-grained descriptions of patients' history, evolution, and treatment. These descriptions are materialized in findings reports, medical orders, as well as in evolution and discharge summaries. In most clinical environments natural language is the main carrier of documentation. Written clinical jargon is commonly characterized by idiosyncratic terminology, a high frequency of highly context-dependent ambiguous expressions (especially acronyms and abbreviations). Violations of spelling and grammar rules are common. The purpose of this work is to map free text from clinical narratives to a domain ontology (SNOMED CT). To this end, natural language processing (NLP) tools will be combined with a heuristic of semantic mapping. The study uses discharge summaries from the Hospital das Clínicas de Porto Alegre, RS, Brazil. Parts of these texts are used for creating a training corpus, using manual annotation supported by active learning technology, used for the training of NLP tools that are used for the identification of parts of speech, the cleansing of "dirty" text passages. Thus it was possible to obtain relatively well-formed and unambiguous noun phrases, heuristics was implemented to semantic mapping between these noun phrases (in Portuguese) and the terms describing the SNOMED CT concepts (English and Spanish) uses the technology of morphosemantic indexing, using a multilingual subword thesaurus, provided by the MorphoSaurus system, the resolution of acronyms, and the identification of named entities (e.g. numbers). In this study, 80 per cent of the summaries are analyzed and manually annotated, resulting in a domain corpus that supports the specialization of the OpenNLP system, mainly following the paradigm of statistical natural language processing (the accuracy of the tagger obtained was 93.67%). Simultaneously, several techniques have been used for validating and improving the subword thesaurus. To this end, the semantic representations of comparable test corpora from the medical domain in English, Spanish, and Portuguese were compared with regard to the relative frequency of semantic identifiers, improving the corpus coverage (2% to Portuguese, and 50% to Spanish). The result was used as an input by a team of lexicon curators, which continuously fix errors and fill gaps in the trilingual thesaurus underlying the MorphoSaurus system. The progress of this work could be objectified using OHSUMED, a standard medical information retrieval benchmark. The mapping of text-encoded clinical information to a domain ontology constitutes an area of high scientific and practical interest due to the need for the analysis of structured data, whereas the clinical information is routinely recorded in a largely unstructured way. In this work the ontology used was SNOMED CT. The evaluation of mapping methodology indicates accuracy of 83.9%.

This thesis discusses possibilities of accelerated 3D scene displaying in a Web browser. In more detail, it deals with WebGL standard and its use in real applications. An application for visualization of volumetric medical data based on JavaScript, WebGL and Three.js library was designed and implemented. Image data are loaded from Google Drive cloud storage. An important part of the application is 3D visualization of the volumetric data based on volume rendering technique called Ray-casting.

The development of new solutions incorporating artificial intelligence (AI) within the medical field is an area of great interest. However, access to comprehensive and diverse datasets is restricted due to the sensitive nature of the data. A potential solution to this is to generatesynthetic datasets based on real medical data. Synthetic data could protect the integrity of the subjects while preserving the inherent information necessary for training AI models and be generated in greater quantity than otherwise available. This thesis project aims to generate reliable clinical trial subject data using a generative adversarial network (GAN). The main data set used is a mock clinical trial dataset consisting of multiple subject visits, however an additional data set containing authentic medical data is also used for better insights into the model’s ability to learn underlying relationships. The thesis also investigates training strategies for simulating the temporal dimension and the missing values in the data. The GAN model used is an altered version of the Conditional Tabular GAN (CTGAN)made to be compatible with the preprocessed clinical trial mock data, and multiple model architectures and number of training epochs are examined. The results show great potential for GAN models on clinical trial datasets, especially for real-life data. One model, trained on the authentic dataset, generates near-perfect synthetic data with respect to column distributions and correlation between columns. The results also show that classification models trained on synthetic data and tested on real data have the potential to match the performance of classification models trained on real data. While the synthetic data replicates the missing values, no definitive conclusion can be drawn regarding the temporal characteristics due to the sparsity of the mock dataset and lack of real correlations in it. Although the results are promising, further experiments on authentic datasets with less sparsity are required.

La gestion électronique des données médicales consiste autant dans le simple traitement automatisé des données personnelles que dans le partage et l'échange de données relatives à la santé. Son encadrement juridique est assuré, à la fois, par les règles communes au traitement automatisé de toutes les données personnelles et par celles spécifiques au traitement des données médicales. Cette gestion, même si elle constitue une source d'économie, engendre des problèmes de protection de la vie privée auxquels le gouvernement français tente de faire face en créant l'un des meilleurs cadres juridiques au monde, en la matière. Mais, de grands chantiers comme celui du dossier médical personnel attendent toujours d'être réalisés et le droit de la santé se voit devancer et entraîner par les progrès technologiques. Le développement de la télésanté bouleverse les relations au sein du colloque singulier entre le soignant et le soigné. L'extension des droits des patients, le partage de responsabilité, l'augmentation du nombre d'intervenants, le secret médical partagé constituent de nouveaux enjeux avec lesquels il faut, désormais compter. Une autre question cruciale est celle posée par le manque d'harmonisation des législations augmentant les risques en cas de partage transfrontalier de données médicales / The electronic management of medical data is as much in the simple automated processing of personal data in the sharing and exchange of health data . Its legal framework is provided both by the common rules to the automated processing of all personal data and those specific to the processing of medical data . This management , even if it is a source of economy, creates protection issues of privacy which the French government tries to cope by creating one of the best legal framework in the world in this field. However , major projects such as the personal health record still waiting to be made and the right to health is seen ahead and lead by technological advances . The development of e-health disrupts relationships within one dialogue between the caregiver and the patient . The extension of the rights of patients , sharing responsibility , increasing the number of players , the shared medical confidentiality pose new challenges with which we must now count. Another crucial question is posed by the lack of harmonization of legislation increasing the risks in cross-border sharing of medical

This thesis deals with CT data segmentation using convolutional neural nets and describes the problem of training with limited training sets. User interaction is suggested as means of improving segmentation quality for the models trained on small training sets and the possibility of using transfer learning is also considered. All of the chosen methods help improve the segmentation quality in comparison with the baseline method, which is the use of automatic data specific segmentation model. The segmentation has improved by tens of percents in Dice score when trained with very small datasets. These methods can be used, for example, to simplify the creation of a new segmentation dataset.

Tato práce popisuje přehled standardních a pokročilých metod používaných k diagnose glaukomu v ranném stádiu. Na základě teoretických poznatků je implementován internetově orientovaný informační systém pro oční lékaře, který má tři hlavní cíle. Prvním cílem je možnost sdílení osobních dat konkrétního pacienta bez nutnosti posílat tato data internetem. Druhým cílem je vytvořit účet pacienta založený na kompletním očním vyšetření. Posledním cílem je aplikovat algoritmus pro registraci intenzitního a barevného fundus obrazu a na jeho základě vytvořit internetově orientovanou tři-dimenzionální vizualizaci optického disku. Tato práce je součásti DAAD spolupráce mezi Ústavem Biomedicínského Inženýrství, Vysokého Učení Technického v Brně, Oční klinikou v Erlangenu a Ústavem Informačních Technologií, Friedrich-Alexander University, Erlangen-Nurnberg.

This diploma thesis is concerned with possibilities of patient data transfer from a pacemaker or implantable cardioverter-defibrillator to physician over GSM. Theoretical part describes options of data transfer in GSM networks, data appropriate for sending and a structure of message is proposed. A device, that is able to send medical data in a form of SMS messages is designed and constructed. The device uses a Freescale MC68HC908GP32 microcontroller, character display with a Hitachi HD44780 controller and a cell phone Sony CMD-J70. The program for microcontroller is written in assembler for HC08. For tabular view of received messages, an application software for PC was created.