Accurate vessel segmentation in ultrasound images using a local-phase-based snake. (May 2018)

• Record Type: Journal Article
• Authors: Ma, Lei; Kiyomatsu, Hidemichi; Nakagawa, Keiichi; Wang, Junchen; Kobayashi, Etsuko; Sakuma, Ichiro
• Is Part Of: Biomedical signal processing and control. Volume 43(2018)
• Page Start: 236
• Page End: 243
• Publication Date: 2018-05
• Subjects: Ultrasound -- Vessel segmentation -- Local phase -- Snake -- Edge detection; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2018.03.002
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A local-phase-based snake is proposed herein to realize accurate vessel segmentation in ultrasound images. In the novel snakes, the image energy is calculated based on the local-phase rather than the conventional gradient. The image energies generated using local phase are identical on all edges regardless of their intensity contrasts. Accurate vessel segmentation can be realized by minimizing the identical local-phase based image energies. The results of the evaluation showed that the local-phase-based-snake performed better than the comparison methods. Abstract: Accurate vessel segmentation in ultrasound images is difficult to realize given the poor image quality of such images. In this study, a local-phase-based snake is proposed to improve the accuracy of vessel segmentation in ultrasound images. In the proposed snake framework, image energy is generated using the local phase of the ultrasound image. By using the intensity-invariant local phase, we can obtain identical image energies on all edges regardless of the strengths of their intensity contrasts. In addition, by minimizing the identical local-phase-based image energy, the snake can be pulled uniformly toward the strong and weak edges, and accurate vessel segmentation in ultrasound images can be achieved. The performance of the proposed segmentation method is evaluated using three types of vessels. We then compare it with the performances of the gradient-based standard snake and snake that uses the balloon model. The evaluation results show that the proposed local-phase-based snake accurately segments vessels from ultrasound images and performs better than the known methods.