Multiscale AM-FM Models and Instantaneous Amplitude Evaluation for Mammographic Density Classification

Breast cancer is the most common cancer in women and the number of incidences keeps rising. Mammographic breast density has been recognized as a very important breast cancer risk and can also mask abnormalities. Information regarding mammographic breast density may be used for planning individualized breast cancer screening and treatment. Thus, breast density is increasingly assessed as the limitations of a onefits-all method of screening and Computer Aided Detection become more apparent. The presented work investigates the use of Amplitude-Modulation Frequency-Modulation (AM-FM) models in the evaluation of multiscale Instantaneous Amplitude (IA) features for the characterization of breast density. AM-FMmodels provide a meaningful and concise method to model digital images. The IA evaluated at different frequency scales is used to capture the relative variations in the breast tissue characteristic to the different breast density classes. Normalized histograms of the IA across the different frequency scales estimated using multiscale Dominant Component Analysis are used to model the breast density classes. Classification of a new mammogram into one of the density categories is achieved using the k-nearest neighbor method and the Euclidean distance metric. The method is evaluated using the Breast Imaging Reporting and Data System on the Medical Image Analysis Society mammographic database and the results are presented and compared to other methods in the literature. The presented method allows breast density classification accuracy reaching over 80%.