Detection of breast cancer in mammography: A neural approach .1. Detection of clustered microcalcifications.

An automatic detection system of breast cancer is being developed in our laboratory. Clustered microcalcifications, stellate images, circumscribed opacities and asymetry of density are investigated. A set of artificial neural networks specialized in the detection of each abnormality, have been defined for this task. This paper present the specialist that performs the detection of clustered microcalcifications. It is a classical three-layer neural network, trained with the Back-Propagation algorithm. The mammogram is analysed by small portions extracted from the image scanned from the left to the right and from the top to the bottom. Success rate is 97 % for clustered microcalcifications regions and 95 % for areas without any microcalcifications.