Background: Breast cancer is a complex disease, both phenotypically and etiologically. Accordingly, the responses to various treatments in the adjuvant setting among individuals vary considerably. There is a demand for tools that can distinguish patients who may benefit or may suffer from particular systemic treatments. We hypothesized that combination of data on genetic biomarkers with data from traditional clinical and pathophysiological (clinicopathologic) factors using traditional Cox model or Support Vector Machine (SVM) method, a new machine learning method, may provide a better tool for prediction of benefits to chemotherapy for the treatment of early breast cancer than using either biomarker or clinicopathologic data alone. Methods: This project included 531 patients from NCIC-CTG MA.5 trial who had data on both clinicopathologic factors, such as age, tumor size, ER status, type of surgery, tumor grade and lymph node involvement, and biomarkers assayed on tissue microarrays (TMAs), including HER2, p53, CA9, MEP21, clusterin, pAKT, COX2 and TOP2A. The Cox model and SVM methods were used to develop prognostic indices for relapse-free or overall survival with either data from TMAs and clinicopathologic assessments alone or their combination. The prognostic indices developed were then examined for their value as predictive classifiers for benefits from CEF treatment. The power of the predictive classifiers derived was evaluated and compared using the bootstrap approach. Results: None of the prognostic indices developed were found to have significant predictive value, although the prognostic index developed using SVM method based on only biomarkers yielded a marginal significant p-value (p=0.0527) for the interaction between classifier and treatment. In accordance with results published previously, the interaction between the classifier developed based on HER2 or TOP2A and treatment was significant (p=0.02 and 0.04 respectively). Comparisons based on the bootstrap approach indicate classifiers developed based on SVM performed better than those based on the Cox model method. Conclusions: Combination of data using biomarkers and clinical-pathological factors, and using either the traditional COX model method or the new machine learning method was not shown to perform better than two single previously known biomarkers in prediction of response to CEF treatment for early breast cancer. / Thesis (Master, Community Health & Epidemiology) -- Queen's University, 2010-04-27 10:16:11.811

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-08-16T18:29:06Z No. of bitstreams: 1 JoberthSilva.pdf: 6383640 bytes, checksum: f18918eb45c49cb426b560e4daddf994 (MD5) / Made available in DSpace on 2017-08-16T18:29:06Z (GMT). No. of bitstreams: 1 JoberthSilva.pdf: 6383640 bytes, checksum: f18918eb45c49cb426b560e4daddf994 (MD5) Previous issue date: 2013-02-20 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Breast cancer is worldwide the most common form of cancer affecting woman, sometimes in their lives, at the proportion of either one to nine or one to thirteen women who reach the age of ninety in the west world (LAURENCE, 2006). Breast cancer is caused by frequent reproduction of cells in various parts of the human body. At certain times, and for reasons yet unknown, some cells begin to reproduce at a higher speed, causing the onset of cellular masses called neoplasias, or tumors, which are new tissue formation, but from pathological origin. This work has proposed a method of automatic detection of masses in digital mammograms, using the Quality Threshold (QT), and the Supporting Vector Machine (MVS). The images processing steps were as follows: firstly, the pre-processing phase took place which consisted of removing the background image, smoothing it with a low pass filter, to increase the degree of contrast, and then, in sequence, accomplishing an enhancement of the Wavelet Transform (WT) by changing their coefficients with a linear function. After the pre-processing phase, came the segmentation with the use of the QT which divided the image in to clusters with pre-defined diameters. Then, the post-processing occurred with the selection of the best candidates to mass formed by the MVS analysis of the shape descriptors. For the extraction phase of texture features the Haralick descriptors and the function correlogram were used. As for the classification stage, the MVS was used again for training, validation of the MVS model and final test. The achieved results were: sensitivity of 92. 31%, specificity of 82.2%, accuracy of 83,53%, a false positive rate per image of 1.12 and an area under a FROC curve of 0.8033. / O câncer de mama é, mundialmente, a forma mais comum de câncer em mulheres afetando, em algum momento suas vidas, aproximadamente uma em cada nove a uma em cada treze mulheres que atingem os noventa anos no mundo ocidental (LAURANCE, 2006). O câncer de mama é ocasionado pela reprodução frequente de células de diversas partes do corpo humano. Em certos momentos e por motivos ainda desconhecidos algumas células começam a se reproduzir com uma velocidade maior, ocasionando o surgimento de massas celulares denominadas de neoplasias ou tumores que são tecidos de formação nova, mas de origem patológica. Neste trabalho foi proposto um método de detecção automática de massas em mamografias digitais usando o Quality Threshold (QT), e a Máquina de Vetores de Suporte (MVS). As etapas de processamento das imagens foram as seguintes: primeiramente veio a fase de pré-processamento que consiste em retirar o fundo da imagem, suavizá-la com um filtro passa-baixa, aumentar a escala de contraste, e na sequencia realizar um realce com a Transformada de Wavelet (WT) através da alteração dos seus coeficientes com uma função linear. Após a fase de pré-processamento vem a seguimentação utilizando o QT que segmenta a imagem em clusters com diâmetros pré-definidos. Em seguida, vem o pós-processamento com a seleção dos melhores candidatos à massa feita através da análise dos descritores de forma pela MVS. Para fase de extração de características de textura foram utiliza os descritores de Haralick e a função correlograma. Já na fase de classificação a MVS novamente foi utilizada para o treinamento, validação do modelo MVS e teste final. Os resultados alcançados foram: sensibilidade de 92,31%, especificidade de 82,2%, Acurácia de 83,53%, uma taxa de falsos positivos por imagem de 1,12 e uma área sob a curva FROC de 0,8033.