A multi-steps segmentation approach for 3D ultrasound images using the combination of 3D-Snake and Level-Set

The segmentation task of 3D ultrasound images has been investigated by a lot of researchers, due to the advantage that it provides to non-invasively detect internal organs and medical conditions. The challenging problems toward the segmentation of 3D ultrasound images are the presence of high speckle noise, inconsistent intensity level and gaps within walls. We propose a new idea by combining 3D-Snake and Level-Set approaches to overcome ultrasound difficulties and maintain high accuracy to segment bleeding regions. The proposed approach starts with a denosing step. The 3D-Snake is resistive to boundaries' discontinuities, and therefore is applied to robustly extract the approximation of the volume of interest. Then, the Level-Set approach is utilized to increase the segmentation accuracy in a few iterations. The parameters of the proposed method are analysed and its segmentation accuracy is evaluated and compared with the Level-Set method. The obtained results validate the superiority of the proposed method.