Ultrasound Image Segmentation

Ultrasound imaging or ultrasonography is an important diagnosis method in medical analysis. It is important to segment out cavities, di erent types of tissues and organs in the ultrasound image for e ective and correct diagnosis. Human experts are very good in segmenting out the required region. But humans lack e ciency when size of dataset increases. The need of high reproducibility and need of increasing e ciency motivates the development of computer-assisted and automated segmentation[1]. These automated procedures segment out di erent regions in medical images by applying di erent types of image segmentation methods. The main disadvantage of ultrasound images is the poor quality of images, which are also a ected by speckle noise. Therefore, in general, many of the image segmentation methods may not be suitable in case of ultrasound images[2]. Watershed segmentation is one of the most e ective methods in complex segmentation problems[3, 4, 1]. The algorithm uses watershed transform applied to images to obtain the segmented regions. However, segmentation of noisy ultrasound image using watershed transform always leads to over-segmentation. In this study some techniques have been studied in order to remove the over-segmentation.