Neste trabalho, propõe-se a utilização da técnica Enhanced ICA Misture Model (EICAMM) para a segmentação automática de nódulos mamários em imagens mamográficas digitais. Com o objetivo de compará-la com outros métodos de segmentação encontrados na literatura correlata, como as técnicas Watershed, Self-Organizing Map (SOM), K-Means e Fuzzy C-Means, utiliza-se a métrica Area Overlap Measure (AOM) ou medida de similaridade de Jaccard, para medir a semelhança entre o resultado obtido na segmentação e o recorte efetuado por um especialista (ground truth). Os resultados obtidos mostram um bom desempenho do modelo EICAMM, que foi a única técnica capaz de detectar massas em regiões de interesse de mama densa. Resultados mais precisos produzidos por tal modelo foram aplicados na elaboração de um módulo classificador de nódulos para um esquema CADx (de Computer-aided Diagnosis) em mamografia digital. O módulo utiliza técnicas de extração e seleção de características e técnicas inteligentes, como Redes Neurais Artificiais, para indicar a existência ou não de nódulos em regiões de interesse, bem como avalia seu contorno/margem, forma e densidade, a fim de indicar a pertinência do achado a um caso maligno ou benigno. Para isso, utiliza-se uma base de regras, criada com o auxílio de um especialista e da combinação de diferentes classificações, conhecida como ensemble, para gerar uma única saída. Testes utilizando várias regiões de interesse selecionadas de duas bases de imagens mamográficas disponíveis resultaram numa precisão média de 46,71% na segmentação dos nódulos pela EICAMM (20,72% melhor que a média das demais técnicas comparadas) e um nível de acerto médio de 80,5% na classificação dos nódulos, o que permite considerar o módulo desenvolvido como uma útil ferramenta para auxílio ao diagnóstico desse tipo de estrutura em esquemas CADx. / This work describes the use of a technique called Enhanced ICA Misture Model (EICAMM) for automated segmentation of breast nodules in digital mammography images. Aiming to compare it with other segmentation methods, like Watershed transformation, Self-organizing Map (SOM), K-Means and Fuzzy C-Means, metrics such as Area Overlap Map (AOM) or Jaccard similarity measure are used in order to measure the similarity between the result from the segmentation and the profile determined by a spectialist (ground truth). Results show a good performance for the EICAMM method, the unique able to detect masses in regions of interest from dense breasts. More accurate results from such a model were applied to the development of a nodules classifier module for a CADx scheme in digital mammography. This module uses techniques for features extraction and selection, and intelligent techniques, as artificial neural networks, to determine the existence or not of a nodule, as well as to evaluate its contour/border, shape and radiographic density, in order to point out its pertinency to a malignant or benign case. With this purpose, a rules database known as ensemble, created with help of a specialist and different classifications combination, is used in order to produce only one output. Tests with several regions of interest selected from two available mammographic images databases have resulted in an average accuracy of 46.71% for nodules segmentation by EICAMM (20.72% better than the average of the other compared techniques), as well as an average accuracy of 80.5% in nodules classification, which allows to consider the developed module as an useful tool in aiding the diagnosis of such a structure in CADx schemes.

This thesis aims to study the effects of endogenous and exogenous hormones and mammographic breast density (BD) on cellular markers in non-cancerous female breast tissue. Women on the waiting list for breast reduction plastic surgery were recruited (n = 79), and randomized to 2 months of hormone therapy or no therapy before surgery. The women had a mammogram and a needle biopsy 2 months before surgery and tissue samples were obtained at the operation. In premenopausal women, estrogen receptor (ER)α levels were associated with age (p = 0.0002), were similar in the follicular and luteal phases of the menstrual cycle and were higher in parous than in nulliparous women (p = 0.009). Current smokers had lower PR levels than non-smokers (p = 0.019). Women on oral contraception had lower ERα (p = 0.048) and PR (p = 0.007) levels than women in the follicular phase. The ERα levels did not differ significantly between postmenopausal estrogen and estrogen-progestogen users, but PR levels were lower among estrogen-progestogen users (p = 0.03). We found lower expression of the genes for decorin and syndecans 1 and 4 in the luteal phase than in the follicular phase, among parous women. Protein levels of the androgen receptor, syndecan-4 and decorin was lower in premenopausal women who were using oral contraceptives (OC) than in those in the follicular phase (p = 0.002 - 0.02), whereas no significant differences between OC use and the luteal phase were found. In premenopausal women, BD was negatively associated with age and body mass index but was similar for the menstrual phases. Breast density was associated with genetic expression of the androgen receptor and remained significant after adjustment for age (rs = 0.56; p = 0.04). After adjustement for age, breast density was also marginally associated with expression of the caspase 3 gene (0.55; 0.053). However, protein levels of caspase 3 was negatively associated (-0.61; 0.03).

Este trabalho apresenta um método computacional que classifica as mamografias no padrão de densidade BI-RADS, visando auxiliar a detecção precoce do câncer de mama, seja essa realizada por análise visual ou por auxílio computadorizado. A classificação das mamografias em bancos padronizados objetiva eliminar conflitos entre laudos mamográficos de diferentes profissionais, bem como quanto à conduta médica a ser seguida. Entretanto, o estabelecimento de bancos feito visualmente e principalmente em períodos diferentes dificulta sua uniformização, proporcionando uma classificação muito subjetiva e relativamente grosseira em conseqüência a grande variação entre e inter observadores. O método desenvolvido permitiu classificar as imagens independentemente da subjetividade própria à observação visual de quem organizou o banco ou da técnica de exposição aos raios X utilizada. Os resultados foram superiores a 92% mesmo para bancos de imagens totalmente diferentes. Esses resultados foram obtidos respeitando-se as possíveis diferenças de interpretações de diversas equipes médicas. Além do estabelecimento de banco de mamografias com limiares entre as composições bem quantificadas, com esta ferramenta, tanto os estagiários poderão ser treinados para classificar as imagens no padrão de densidades do BI-RADS, respeitando as particularidades locais, quanto os resultados dos CAD poderão ser comparados. / This thesis presents a computational method that classifies the mammography into the composition of the breast tissue density patterns described in the BI-RADS protocol, intended to help in the early detection of breast cancer, either if this detection happens to be realized by visual analysis or by computerized support. The classification of the mammography in standardized database intends to eliminate issues between mammography awards of distinct professionals and the correct medical conduct to be followed. However, the determination of database only visually, especially in different periods, difficult it\'s to standardize, causing an extremely subjective classification and relatively superficial in consequence of the large inter-and intraobserver variability. The method allows classifying the images independently of the subjective quality of the visual analysis from who organized the database or from the technique of the exposition to X-ray employed. The results were superior of 92% even to database totally distinct. These results were obtained respecting eventual differences of interpretation from several medical groups. Beside the establishment of mammography database with thresholding between the well quantified categories, this methodology will consent to probationers to be trained for classify the images according to the composition of the breast tissue density patterns described in the BI-RADS, respecting its local particularity. Likewise, with this methodology, the results from CAD would be compared.

No description available.