Level sets driven by adaptive hybrid region-based energy for medical image segmentation

Abstract

Medical image segmentation has a great significance for medical diagnosis. In this article, a new level set method (LSM) driven by adaptive hybrid region-based energy is proposed to achieve accurate medical image segmentation. First, new median region intensity descriptions are computed using the filtered input image and combined with traditional mean region intensity descriptions to design a novel global region-based signed pressure force (GRSPF). Then, the global region-based energy is defined using this GRSPF. Second, a similar new local region-based SPF (LRSPF) is also designed and the local region-based energy is defined using this LRSPF, which enhances the model’s versatility. Furthermore, an adaptive weight for controlling the roles of the global and local region-based energies is introduced to construct the hybrid region-based energy, which drives the level set more appropriately. Segmentation results for medical images show that the proposed LSM can segment medical images more accurately than existing LSMs.