Hybrid segmentation of microcalcification in mammogram images by using Fractional Order Darwinian Particle Swarm Optimization and Mathematical Morphology Methods

Image segmentation is a process of extracting desired area of microcalcification on mammogram images. In mammograms, microcalcification can be seen as early sign of breast cancer. The detection of microcalcification is difficult to be identified due to low contrast and existing noise between microcalcification and surrounding tissue. A technique that has been widely used in segmentation is Particle Swarm Optimization (PSO). However, a general problem with PSO and other optimization algorithms is to get stuck in an optimum local stage where it can do well on certain images but can fail on another. Thus, Fractional Order Darwinian Particle Swarm Optimization (FODPSO) is used to segment microcalcification and distinguish from the background tissue of mammogram images. In general, FODPSO has a high fitness value to generate an efficient segmentation, but there are still some images that come out with unwanted regions of microcalcification. Therefore, an improved version of the FODPSO by using the Mathematical Morphology (MM) method is conducted to enhance the segmentation results. Morphology involves theory of analyzing the shape and texture of the image. The role of MM is used to obtain smooth shape and remove unwanted pixels of microcalcification. The algorithms are tested on 30 mammogram images which consists of microcalcification. The performance of improved FODPSO is evaluated by measuring the accuracy and sensitivity of the segmentation result. The evaluation used the percentage relative error of area and compared the result between method and expert. The accuracy and sensitivity of the segmentation results are 93.30% and 83.33% respectively. Hence, it is proven that the improved FODPSO method has the ability to segment the microcalcification on mammogram images and it can give a future reference.