Aplicação de um método computacional no diagnóstico precoce do câncer de próstata baseado em reconhecimento de padrões proteómicos

• Ano de defesa: 2017
• Autor(a) principal: Montes, Elzenir de Araujo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: UEMA; Campus São Luis Centro de Ciências Tecnológicas – CCT; PROGRAMA DE PÓS-GRADUAÇÃO EM ENGENHARIA DA COMPUTAÇÃO E SISTEMAS - PECS
• Palavras-chave em Português: Sinais proteômicos; Análise de componentes independentes; Máxima relevância; Mínima redundância; Máquina de vetores de suporte; Análise discriminante linear; Câncer de próstata; Algoritmo FastICA; Biomarcadores; Proteômica; Aprendizado de máquinas; Diagnóstico de câncer de próstata; Proteomic signals; Independent component analysis; Maximum relevance; Minimum redundancy; Support vector machine; Linear discriminant analysis; Prostate cancer; FastICA algorithm; Biomarkers; Proteomics; Machine learning; Ciência da Computação
• Link de acesso: http://repositorio.uema.br/handle/123456789/461

Prostate cˆancer in its early stages has a fairly steady evolution, so much so that most patients do not show symptoms, and when they appear they are confused with benign prostate growth. This work proposes the application of a set of computational techniques to compose a new method of early diagnosis of prostate cancer, based on recognition of proteomic patterns. The method has basically three steps. The first step is performed by the Independent Component Analysis (ICA) technique through the FastICA algorithm, in order to extract the characteristics of the proteomic signals. The second step aimed at reducing the set of characteristics, and with this the computational cost was used the technique Maximum Relevance and Minimum Redundancy (mRMR). In the third step, two classifiers were used to compare the results between them and to decide on the best set of techniques to be used in the early diagnosis of prostate cancer, Supporting Vector Machine (SVM) and Linear Discriminant Analysis (LDA) . Thus, the results obtained with set of techniques (ICA ⇒ mRMR ⇒ SVM) were satisfactory, but it was making use of the set (ICA ⇒ mRMR ⇒ LDA) that the best results were achieved From a vector of 77 characteristics, the LDA classifier obtained an excellent response in the classification phase, obtaining accuracy, specificity and sensitivity respectively of 100 %, 100 % and 100 %.