Spatial density modeling for discirminating between benign and malignant microcalcification lesions

Accurate diagnosis of microcalcification (MC) lesions in mammograms is an important but challenging clinical task in early cancer detection. In this work, we investigate how to extract salient and robust quantitative features for discriminating between benign and malignant cases in the presence of inaccuracy in MC detection. We propose to use a spatial density function (SDF) to characterize the spatial distribution of the MCs in a cluster, aimed to better accommodate the potential inaccuracy in the detected MCs. We demonstrate this approach on a set of commonly used features for clustered MCs. The proposed approach was tested on a set of 640 cases. The results show that the SDF features are robust to variations in MC detection while achieving better class separation.