CAD Method for Microcalcification Detection: Independent of Sensor and Resolution

摘要

The aims of this work are to explore the feasibility of developing a new class of computer assisted diagnostic (CAD) methods for microcalification cluster (MCC) detection for breast cancer screening using digital mammography. The objectives are to achieve: (a) improved CAD performance that is significantly more robust for large image databases, and (b) an adaptive CAD method that is independent of the digital sensor resolution and gray scale characteristics; for the first time. This report includes 3 sections, (I). Summary of the work in first year, which includes data base collection and truth file establishment for different sensors, preprocessing for breast area segmentation, and basic algorithm design and optimization, (2) Summary of the work in second year, which includes algorithm design and modular optimization for enhancement, segmentation, feature extraction and classification. (3). Whole system optimization and evaluation, which includes a design, optimization and evaluation of a successful MCCs detection system.