Multiresolution detection of spiculated lesions in digital mammograms

In this paper, we present a novel multiresolution scheme for the detection of spiculated lesions in digital mammograms, First, a multiresolution repre sentation of the original mammogram is obtained using a linear phase nonsep arable two-dimensional (2-D) wavelet transform. A set of features is then e xtracted at each resolution in the wavelet pyramid for every pixels This ap proach addresses the difficulty of predetermining the neighborhood size for feature extraction to characterize objects that may appear in different si zes. Detection is performed from the coarsest resolution to the finest reso lution using a binary tree classifier. This top-down approach requires less computation by starting with the least amount of data and propagating dete ction results to finer resolutions. Experimental results using the MIAS ima ge database have shown that this algorithm is capable of detecting spiculat ed lesions of very different sizes at low false positive rates.