A Stroke can affect different parts of the human body depending on the area of brain effected; our research focuses on upper limb motor dysfunction for stroke patients. In current practice, ordinal scale systems are used for conducting physical assessment of upper limb impairment. The reliability of these assessments is questionable, since their coarse ratings cannot reliably distinguish between the different levels of performance. This thesis describes the design, implementation and evaluation of a novel system to facilitate stroke diagnosis which relies on data collected with an innovative KINARM robotic tool. This robotic tool allows for an objective quantification of motor function and performance assessment for stroke patients. The main methodology for the research is Case Based Reasoning (CBR) - an effective paradigm of artificial intelligence that relies on the principle that a new problem is solved by observing similar, previously encountered problems and adapting their known solutions. A CBR system was designed and implemented for a repository of stroke subjects who had an explicit diagnosis and prognosis. For a new stroke patient, whose diagnosis was yet to be confirmed and who had an indefinite prognosis, the CBR model was effectively used to retrieve analogous cases of previous stroke patients. These similar cases provide useful information to the clinicians, facilitating them in reaching a potential solution for stroke diagnosis and also a means to validate other imaging tests and clinical assessments to confirm the diagnosis and prognosis. / Thesis (Master, Computing) -- Queen's University, 2008-09-27 11:14:04.85

Made available in DSpace on 2016-06-02T19:05:58Z (GMT). No. of bitstreams: 1 4486.pdf: 3511223 bytes, checksum: 3f8c09a3cb87230a2ac0f6706ea07944 (MD5) Previous issue date: 2012-08-08 / Financiadora de Estudos e Projetos / The mining association rules are an important task in data mining. Traditional algorithms of mining association rules are based only on the database items, providing a very specific knowledge. This specificity may not be advantageous, because the users normally need more general, interesting and understandable knowledge. In this sense, there are approaches working in order to obtain association rules with items belonging to any level of a taxonomic structure. In the crisp contexts taxonomies are used in different steps of the mining process. When the objective is the generalization they are used, mainly, in the pre-processing or post-processing stages. On the other hand, in the fuzzy context, fuzzy taxonomies are used, mainly, in the pre-processing step, during the generating extended transactions. A great problem of these transactions is related to the huge amount of candidates and rules. Beyond that, the inclusion of ancestors ends up generating redundancy problems. Besides, it is possible to see that many works have directed efforts for the question of mining fuzzy rules, exploring linguistic terms, but few approaches have been proposed for explore new steps of mining process. In this sense, this paper proposes the Context FOntGAR algorithm, a new algorithm for mining generalized association rules under all levels of fuzzy ontologies composed by specialization/generalization degrees varying in the interval [0,1]. In order to obtain more semantic enrichment, the rules may be composed by similarity relations, which are represented at the fuzzy ontologies in different contexts. In this work the generalization is done during the post-processing step. Other relevant points of this paper are the specification of a new approach of generalization; including a new grouping rules treatment, and a new and efficient way for calculating both support and confidence of generalized rules. / Algoritmos tradicionais de associação se caracterizam por utilizar apenas itens contidos na base de dados, proporcionando um conhecimento muito específico. No entanto, essa especificidade nem sempre é vantajosa, pois normalmente os usuários finais necessitam de padrões mais gerais, e de fácil compreensão. Nesse sentido, existem abordagens que não se limitam somente aos itens da base, e trabalham com o objetivo de minerar regras (generalizadas) com itens presentes em qualquer nível de estruturas taxonômicas. Taxonomias podem ser utilizadas em diferentes etapas do processo de mineração. A literatura mostra que, em contextos crisp, essas estruturas são utilizadas tanto em etapa de pré-processamento, quanto em etapa de pós-processamento, e que em domínios fuzzy, a utilização ocorre somente na etapa de pré-processamento, durante a geração de transações estendidas. Além do viés de utilização de transações estendidas, que podem levar a geração de um volume de regras superior ao caso tradicional, é possível notar que, em domínios nebulosos, as pesquisas dão enfoque apenas à mineração de regras fuzzy, deixando de lado a exploração de diferentes graus de especialização/generalização em taxonomias. Nesse sentido, este trabalho propõem o algoritmo FOntGAR, um novo algoritmo para mineração de regras de associação generalizadas com itens presentes em qualquer nível de ontologias compostas por graus de especialização/generalização variando no intervalo [0,1] (ontologias de conceitos fuzzy), em etapa de pós-processamento. Objetivando obter maior enriquecimento semântico, as regras geradas pelo algoritmo também podem possuir relações de similaridade, de acordo com contextos pré-definidos. Outros pontos relevantes são a especificação de uma nova abordagem de generalização (incluindo um novo tratamento de agrupamento das regras), e um novo e eficiente método para calcular o suporte estendido das regras generalizadas durante a etapa mencionada.