Can multiscale fractal dimension analysis of screening mammograms detect the architectural distortion of the breast?

Retrospective analysis of false negative screening mammograms shows the high prevalence of the architectural distortion(AD). For an efficient, early manifestation, and higher prognosis rate of breast cancer, this paper examines whether the multiscale fractal dimension(FD) measurements of mammograms is more efficient than single FD measurement for the detection of architectural distortions in screening mammograms. The proposed AD detection approach is based on two key steps: the application of two-dimensional empirical mode decomposition(2D-EMD) algorithm to generate multiscale representation of the mammograms; the second step is measuring the fractal dimension from the multiresolution representation of the mammogram. The proposed approach has been tested using a dataset of mammographic regions of interest extracted from the Mammographic Image Analysis Society (MIAS) database. The ROI dataset contains 19 abnormal regions related to architectural distortions and 207 regions representing normal breast parenchyma. Achieving a classification accuracy of 0.917 using the multiscale fractal dimension measurements demonstrates a very promising performance and supports further examination and extension of the proposed approach.