ABSTRACT
This chapter provides new wavelet-based techniques for the automated detection and semi-automated detection of certain early signs of cancer. It is concerned with the automated detection of clusters of microcalcifications. The detection of microcalcification clusters has several features in common with other problems of feature classification and automated detection. Signs of breast cancer are typically categorized by radiologist in three categories: clusters of microcalcifications, stellate lesions, and circumscribed lesions. The number of mammograms generated daily is large and therefore it is very desirable to develop image processing tools which facilitate the handling of mammograms and aid the radiologist in diagnosis. The chapter presents a general strategy for constructing algorithms for extracting microcalcification clusters and aims to distinguish between two types of algorithms. The first are autonomous and do not involve a computer operator. The second are interactive and allow decisions to be made to improve parameter settings which depend on the image.
