Mammographic lesion detection based on improved concentric morphology model

This paper presents a novel lesion detection algorithm based on the layer structuring hypothesis where different layers were obtained with different thresholds adaptively determined from the histogram of the mammogram. Highly suspicious lesion regions were obtained upon selection procedures based on morphological features and the Single Concentric Layers (SCL) Criterion. A total of 170 mammograms were selected from the MIAS dataset for evaluations of the proposed algorithms. To evaluate performance, FROC analysis was performed. The results indicate that our method is of potential application as an aid to the radiologists in mammograms interpretation.