A shape cognitron neural network for breast cancer detection
San-Kan Lee, Pau‐Choo Chung, Chein‐I Chang, Chien-Shun Lo, Tain Lee, Giu‐Cheng Hsu, Chin-Wen Yang · 2003
A Neocognitron-like neural network built with universal feature planes, called shape cognitron (S-cognitron) is introduced to classify clustered microcalcifications (MCCs). The S-cognitron is composed of two modules. The first module consists of (a) a shape orientation layer, to convert first-order shape orientations into numeric values, and (b) a complex layer to extract second-order shape features. Following is a 3-D figure layer to extract the shape curvatures. It is then followed by a second module made up of a feature formation layer and a probabilistic neural network (PNN)-based classification layer, to construct "potential" high-order shape features and perform the classification. Experimental results show the promise of the system.