Gray Scale Features Extracted from the Breast Profile

An alternative approach to feature extraction is based on the gray scales that comprise a breast profile. A method for the construction of a segmented breast profile is described in [19]. A number of gray-scale features are extracted from the mammograms, including co-occurrence matrix based features, Fourier Transform-spectral energy based features, Law's mask features, discrete wavelet transform features, statistical features, circularity (shape) feature and fractal dimension feature. These are described in detail in [26]. Altogether this gives a set of 316 features FBP316 = (fi, f2,..., f316) that are extracted from the breast profile to characterize a mammogram. In addition, the application of principal component analysis (PCA) (a mechanism for dimensionality reduction) results in a 26-dimensional feature space FBP26 = (f1, f2,..., f26). Both features sets {FBP316, FBP26) are evaluated as mapping features in the learning of the expert contrast enhancement.

0 0

Post a comment