Efeito dos dados clínicos e achados radiológicos na predição de laudos mamográficos

• Ano de defesa: 2017
• Autor(a) principal: OLIVEIRA, Ana Gabriela Caldas
• Orientador(a): SANTOS, Alcione Miranda dos
• Banca de defesa: PALACIO, Glaucia Andrade e Silva , MARTINS, Marília da Glória , THOMAZ, Erika Bárbara Abreu Fonseca , BATISTA, Rosângela Fernandes Lucena
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Maranhão
• Programa de Pós-Graduação: PROGRAMA DE PÓS-GRADUAÇÃO EM SAÚDE COLETIVA/CCBS
• Departamento: DEPARTAMENTO DE SAÚDE PÚBLICA/CCBS
• País: Brasil
• Palavras-chave em Português: Câncer de mama; BI-RADS®; Árvore de classificação; Redes neurais artificiais
• Palavras-chave em Inglês: Breast cancer; Classification tree; Artificial Neural Network
• Área do conhecimento CNPq: Cancerologia; Radiologia Médica
• Link de acesso: https://tedebc.ufma.br/jspui/handle/tede/tede/2076

Breast cancer is the second most common type of cancer in the population and the most common among women. Mammography remains the only method capable of reducing mortality from the disease. The main objective of this study was to evaluate the predictive factors and importance of the clínical variables for mammography reports according to the BI-RADS® (Breast Imaging Reporting and Data System) classification in women who performed mammograms in the city of São Luís (MA). A cross-sectional study was carried out between June 2014 and October 2015, with women receiving care in the public and private network. Demographic and socioeconomic characteristics, life and reproductive habits, as well as the mammographic report according to BIRADS® classification were evaluated. In the first article, the chi-square test and a classification tree type model were used to identify the predictors of BI-RADS® 0 occurrence. Factors related to the high rate of inconclusive mammograms were found to be premenopausal, irregular adhesion to mammography screening, age between 40 and 49 years, and BMI (Body Mass Index) ≥ 25 kg/m2. In the second article, the objective was to evaluate the importance of age, clinical and reproductive characteristics in BI-RADS® classification using ANN (artificial neural network). The overall accuracy of ANN constructed only with characteristics of the mammographic report was 88.5% in the training phase and 78.5% in the test. With the association of the clinical variables, there was an improvement in the overall accuracy of the network, passing to 90% training phase and 83.5% in the test. This improvement was more evident in the BI-RADS® 0 classification (68.6%) and BI-RADS® 4 and 5 (66.7%). The accuracy in classifying BI-RADS® 1, 2 and 3 patients was similar in both networks, around 95%. Regarding the importance of the variables in the construction of ANN, the presence of nodule followed by the presence of calcification were the most important variables in both ANN; Age, being in HRT (hormone replacement therapy), menopause and early menarche, the most important clinical and reproductive variables. The results found point to the high frequency of mammograms classified BI-RADS® 0 in women aged between 40 and 59 years and with a BMI ≥ 25 kg / m2, in addition it was concluded that age, clínical and radiological variables can help the Radiologist for classification of mammography reports.