Optimization of active-contour model parameters using genetic algorithms: segmentation of breast lesions in mammograms

Abstract

Listen

Translate article

Genetic algorithms (GA's) were used to find optimal sets of parameters for an active contour model (ACM) algorithm that segments breast lesions in mammography images. These parameters, which are typically determined empirically, are used in an energy function that is minimized by the ACM algorithm when producing a segmentation contour. Using manually segmented contours supplied by experienced radiologists, GA techniques were used to vary the parameter values until the contours produced by the ACM algorithm closely matched those of the radiologists.