Adaptive wavelet analysis and classification of mammographic microcalcification

The study of mammographic microcalcifications using wavelets has been shown to provide useful information for the diagnosis of breast cancer. In most analyses of images using wavelets, the samples of the image are assumed to be the scaling function expansion coefficients at a certain ''scale''. For large values of the ''scale'', the approximation works since the scaling function approximates a delta function. In this investigation we adapt our scaling function from a class of possible scaling functions. The approximation error at a certain ''scale'' is minimized using a genetic algorithm. The results indicate a promising procedure for selection of optimal wavelets for the subsequent classification of mammographic microcalcifications.