Detecting region-of-interest (ROI) in digital mammogram by using morphological bandpass filter

摘要

Microcalcifications are an early sign of breast cancer and appear as isolated bright spots in mammogram images, which are difficult to detect due to their small size. The paper introduces a morphological bandpass filter (MBF) to detect microcalcifications; it is implemented by opening the original image twice with two different structure elements, and subtracting one opened image using another one that can decompose the image into details of interest in the image domain where microcalcifications tend to appear. Series of MBFs are tuned for the detection task, and binary images containing microcalcification regions-of-interest (ROI) are obtained. Experimental results show that the proposed method with these bandpass filters can recognize ROI with a true positive of 93.07% and a false positive of 4.34%. Compared to the well-known discrete wavelet transform (DWT) method, this method is more accurate in position and size of microcalcifications.