Análise da segmentação de lesões mamárias por watershed com variação de elementos estruturantes

• Autor(a) principal: Goes, Bruno Augusto Machado
• Data de Publicação: 2021
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/33473
• topic: Câncer de mama; Mamografia; Segmentação; Watershed; CNPQ::ENGENHARIAS::ENGENHARIA BIOMEDICA

Breast cancer is the main type of cancer diagnosed worldwide, with one in six women diagnosed with the disease dying. To improve the prognosis and survival of patients is necessary to perform the early detection through mammography. Mammography is a radiographic exam that allows to note mammary lesions as masses and calcifications. Using digital mammographic images processing, it’s possible to improve visualization of structures and facilitate diagnosis, using noise-reduction filters, contrast enhancement and segmentation. Segmentation is the division of the image in regions of interest (lesions) and background. One of the techniques in segmentation is watershed segmentation, that is a topographic analysis of the image and object’s boundaries extraction. Watershed demands using markers, that limits the regions where processing will be applied, avoiding issues as over-segmentation. The markers can be defined using morphological operators, which uses structuring elements to identify interest pixels. Therefore, the aim of this work is to evaluate watershed segmentation of mammographic lesions, by changing the pre-processing and structuring elements. There were applied disk and diamond-shaped structuring elements, with 25, 50 and 100 pixels as size, in addition to noisereduction Wiener filter and contrast enhancement with Wavelet transform and contrast limited adaptative histogram equalization (CLAHE). The results were evaluated by using accuracy, ROC curve (Receiver Operating Characteristic Curve) and visual analysis of the images, comparing the results with the original image. As result, it was possible to identify the relationship between size and pre-processing with the number of detected regions. Moreover, small structuring elements are more effective in watershed segmentation, requiring research with different noise reduction methods and marker definition algorithms. Finally, the results allowed to establish how breast density and shape, size and type of lesion interfere in watershed segmentation results.