Diagnostic Classification of Digital Mammograms by Wavelet-Based Spectral Tools: A Comparative Study

摘要

The aim of this paper is to present results from a comparative investigation into the diagnostic performance of several wavelet-based estimators of scaling, some from published literature and some newly proposed. These estimators are evaluated based on their ability to classify digitized mammogram images from a clinical database, for which the true disease status is known by biopsy. We found that Abry-Veitch and modified weighted Theil-type estimators provided the best classification rates, while the standard wavelet-based OLS estimator performed worst. The results are robust with respect to choice of wavelets (Haar wavelet being an exception) and are of potential clinical value. The diagnostic is based on the properties of image backgrounds (which is an unused diagnostic modality in Mammograms) and the best correct classification rates achieve 90%, varying slightly with the choice of basis, levels used, and size of training set.