Computer Aided Diagnosis in Digital Mammograms: Detection of Microcalcifications by Meta Heuristic Algorithms

This research applies the meta-heuristic methods such as Ant Colony Optimization (ACO) and Genetic Algorithm (GA) for identification of suspicious region in mammograms. The proposed method uses the asymmetry principle (bilateral subtraction): Strong structural asymmetries between corresponding regions in the left and right breast are taken as evidence for the possible presence of microcalcification in that region. Bilateral subtraction is achieved in two steps. First, the mammogram images are enhanced using median filter, pectoral muscle region is removed and the border of the mammogram is detected for both left and right images from the binary image. The enhancement technique is evaluated by signal to noise ratios. Further GA is applied to enhance the detected border. The figure of merit is calculated to identify whether the detected border is exact or not. And the nipple position is identified for both left and right images using GA and ACO, and their performance is studied. Second, using the border points and nipple position as the reference the mammogram images are aligned and subtracted to extract the suspicious region. Results obtained with a set of mammograms indicate that this method can improve the sensitivity and reliability of the systems for automated detection of breast tumors i.e. microcalcification. The algorithms are tested on 161 pairs of digitized mammograms from Mammographic Image Analysis Society (MIAS) database. A Free-Response Receiver Operating Characteristic (FROC) curve is generated for the mean value of the detection rate for all the 161 pairs of mammograms in MIAS database, to evaluate the performance of the proposed method.