Texture feature extraction for tumor detection in mammographic images
M. Sameti, Rabab Kreidieh Ward, Branko Palcic, Jacqueline Morgan-Parkes · 2002
A set of texture features are extracted from segmented regions of digitized mammograms for classification of masses from normal regions. The mass detection algorithm consists of two steps. In the first step, the algorithm employs a segmentation method based on the fuzzy sets theory to divide a mammogram into different regions and produces region(s) of mass candidates. In the second step, discrete texture features are calculated for the area of each mass candidate. Two of those feature were sufficient to produce a 94% true-positive detection rate with a low 0.24 false-positives per image for a data set of 35 mammograms with a malignant mass in each.