Segmenting Masses in Ultrasound Images by Using Seed Based Region Growing and Mathematical Morphology

Ultrasound screening is one of the imaging modalities that can be used in early detection of breast cancer. However, most of the ultrasound images produced artifacts such as attenuation, speckle, shadows, and signal dropout. These artifacts can cause missing boundaries due to the orientation dependence of acquisition. Moreover, the difficulty also occurs as the contrast is frequently low between foreground (areas of interest) and background. Hence, it can give difficulties for a radiologist in interpreting the important information from the image. In order to overcome these problems, a good segmentation technique is required. In this study, Seed Based Region Growing (SBRG) is used as the segmentation method in identifying masses in ultrasound images. The result from SBRG method is enhanced by using Mathematical Morphology technique during post-processing phase to obtain a smooth shape and a clear boundary of masses. Mathematical Morphology involves a theory for the analysis of shape and spatial structures. These methods are tested on 30 ultrasound images confirmed by a radiologist to contain masses. The average percentage area between the methods and the actual area traced by the radiologist is then computed. Experimental results show that the methods have segmented the masses with a percentage difference of 8.46%. This implies that the methods have successfully segmented the masses of ultrasound images with an accuracy of 91.54%.