Caracterización de Patrones Anormales en Mamografías

Abstract. Computer-guided image interpretation is an extensive research area whose main purpose is to provide tools to support decision-making, for which a large number of automatic techniques have been proposed, such as, feature extraction, pattern recognition, image processing, machine learning, among others. In breast cancer, the results obtained at this area, they have led to the development of diagnostic support systems, which have even been approved by the FDA (Federal Drug Administration). However, the use of those systems is not widely extended in clinic scenarios, mainly because their performance is unstable and poorly reproducible. This is due to the high variability of the abnormal patterns associated with this neoplasia. This thesis addresses the main problem associated with the characterization and interpretation of breast masses and architectural distortion, mammographic findings directly related to the presence of breast cancer with higher variability in their form, size and location. This document introduces the design, implementation and evaluation of strategies to characterize abnormal patterns and to improve the mammographic interpretation during the diagnosis process. The herein proposed strategies allow to characterize visual patterns of these lesions and the relationship between them to infer their clinical significance according to BI-RADS (Breast Imaging Reporting and Data System), a radiologic tool used for mammographic evaluation and reporting. The obtained results outperform some obtained by methods reported in the literature both tasks classification and interpretation of masses and architectural distortion, respectively, demonstrating the effectiveness and versatility of the proposed strategies.

• Autores: Narváez Espinoza, Fabián Rodrigo
• Tipo de recurso: Doctoral thesis
• Fecha de publicación: 2016
• Institución: Universidad Nacional de Colombia
• Repositorio: Universidad Nacional de Colombia
• Idioma: spa
• OAI Identifier: oai:repositorio.unal.edu.co:unal/59251
• Acceso en línea: https://repositorio.unal.edu.co/handle/unal/59251; http://bdigital.unal.edu.co/56605/
• Palabra clave: 6 Tecnología (ciencias aplicadas) / Technology; 61 Ciencias médicas; Medicina / Medicine and health; Breast Masses; BI-RADS; Architectural distortion; Patterns Recognition; Multiresolution image analysis; Content based Image Retrieval; Cáncer de mama; Nódulos de Mama; Distorsión Arquitectural; Reconocimiento de Patrones; Análisis Multiresolución; Recuperación de Imágenes basado en Contenido
• Rights: openAccess
• License: Atribución-NoComercial 4.0 Internacional