• Entrar
    Ver item 
    •   Página inicial
    • BIBLIOTECA DIGITAL: Teses & Dissertações
    • 40001016034P5 Programa de Pós-Graduação em Informática
    • Dissertações
    • Ver item
    •   Página inicial
    • BIBLIOTECA DIGITAL: Teses & Dissertações
    • 40001016034P5 Programa de Pós-Graduação em Informática
    • Dissertações
    • Ver item
    JavaScript is disabled for your browser. Some features of this site may not work without it.

    Automatic detection for acute lymphoid leukemia images based non local region approach

    Thumbnail
    Visualizar/Abrir
    R - D - JOAO FELIPE LOPES DE SUS.pdf (6.933Mb)
    Data
    2018
    Autor
    Sus, João Felipe Lopes de
    Metadata
    Mostrar registro completo
    Resumo
    Resumo: A leucemia linfoide aguda (LLA) é o tipo de câncer mais comum a se manifestar na infância, apesar de apresentar rápida evolução em seu quadro clínico a LLA possui relativamente baixa mortalidade quando identificada e tratada em seu estágio inicial. Como o diagnóstico de LLA é feito por médicos hematologistas com base na análise microscópica de lâminas contendo amostras de sangue periférico, o que pode ser considerado um trabalho lento e cansativo impactando no desempenho do médico, o desenvolvimento de ferramentas que auxiliem neste processo é uma necessidade real. A proposta deste trabalho é apresentar um algoritmo capaz de segmentar os leucócitos existentes, extrair e selecionar características gerando uma representação compacta e por fim utilizar um classificador para diferenciar imagens de sangue periférico de pacientes saudáveis de pacientes portadores de LLA. A base de imagens ALL_IDB foi escolhida para ser utilizada por ser uma base de domínio público e também utilizada em outros trabalhos permitindo comparações precisas, e apresentar diversas dificuldades encontradas no trabalho com imagens provenientes de microscópio, como diferentes tipos de iluminação e zoom. Das 108 imagens utilizadas nos testes 107 foram classificadas corretamente, resultando em uma acurácia de 0,99 sendo este valor maior que o melhor trabalho encontrado na literatura atual, mesmo o único caso classificado erroneamente foi um falso positivo o que no contexto da aplicação é menos grave do que um falso negativo.
     
    Abstract: Acute lymphoblastic leukemia (ALL) is the most common type of cancer in childhood. Despite presenting a rapid evolution in its clinical condition, ALL has a relatively low mortality when identified and treated in its initial stage. Due to the fact that the ALL diagnosis is made by hematologists based on the microscopic analysis of the peripheral blood smear slices, which can be considered a tedious and tiring work, impacting on the doctor's performance, the development of tools that would help in this process is a real necessity. Thus, the purpose of this work is to present an algorithm capable of segmenting the existing leukocytes from blood smear images, extracting and selecting the most representative features, generating a compact representation, so as to finally use a classifier to differentiate the peripheral blood smear images of healthy patients from patients with ALL. The ALL_IDB image base was chosen for being a public domain base and also used in other works, thus allowing accurate comparisons, as well as revealing several difficulties that are faced when working with microscopic images, such as different types of lighting and distinct zoom levels. The final results were expressive and reached an accuracy of 0.99, where, from the 108 images used in the tests, 107 were classified correctly. This result is higher than the best one found in the latest literature, and the only image classified as being wrong was a false positive which in the application context is not the worse case scenario.
     
    URI
    https://hdl.handle.net/1884/58511
    Collections
    • Dissertações [255]

    DSpace software copyright © 2002-2022  LYRASIS
    Entre em contato | Deixe sua opinião
    Theme by 
    Atmire NV
     

     

    Navegar

    Todo o repositórioComunidades e ColeçõesPor data do documentoAutoresTítulosAssuntosTipoEsta coleçãoPor data do documentoAutoresTítulosAssuntosTipo

    Minha conta

    EntrarCadastro

    Estatística

    Ver as estatísticas de uso

    DSpace software copyright © 2002-2022  LYRASIS
    Entre em contato | Deixe sua opinião
    Theme by 
    Atmire NV