| This thesis presents a new fuzzy neural network (FNN) approach to detect malignant mass lesions on mammograms. Mammograms were obtained from the digital database for screening mammography (DDSM) at the University of South Florida. Six-hundred-seventy regions of interest (ROIs) were extracted from 100 mammograms and are randomly divided into two groups: training and testing groups. Entropy, uniformity, contrast, and maximum co-occurrence matrix elements are calculated at sizes of 256 x 256 and 768 x 768 pixels. The differences of these features (feature differences) at these two image sizes are computed for each feature. These feature differences are discriminant in differentiating between malignant masses and normal tissues regardless of lesion shape, size, and subtlety. After training, the FNN can correctly detect all malignant masses on mammograms in the testing group. The true-positive fraction (TPF) is 0.92 when the number of false positives (FP) is 1.33 per mammogram and 1.0 when the FP is 2.15 per mammogram. |
