Breast asymmetry analysis employing tree-structured wavelet transform

Abstract

Statistically distributed developmental asymmetries appearing in paired body structures such as breasts in women are usually related to unhealthy biological conditions. In particular, there is evidence that breast cancer patients show more breast dimensional asymmetries and larger breasts than age-matched healthy women. It was also recently reported the application of Gabor nonorthogonal wavelets filtering for analyzing asymmetries of the directional structures appearing in mammograms corresponding to healthy and breast cancer patients. In this paper the Tree-Structured Wavelet Transform, which uses orthogonal wavelet bases, is applied for the first time in order to quantify texture asymmetries between left- and right- mammograms in women. In order to assess the suitability of the method, it was applied to characterize breast asymmetries of mammograms corresponding to a small population of 63 healthy women with ages ranging between 30 and 70 years. The obtained results justify further improvements of this method and its application for correlating asymmetries and the predisposition to breast cancer.