Fast Segmentation for Medical Ultrasound Image Based on Parametric Level Set Active Contour Model

摘要

In this paper, an active contour model based on parametric level set is proposed to solve the problem of fast segmentation of lesions during minimally invasive surgery. The parameterized level set is embedded into the classical non-parametric geometric active contour model of localGaussian distribution fitting, so that not only the topology change is naturally merged into the evolution of the curve, but also the dimension isn't increased. The parametric level set is determined by the parameter vector rather than the traditional signed distance function, so there isno need to add regularization item in energy functional, nor does it need to be reinitialized. The computing cost of the model is reduced and the computing speed is improved. And the dense initialization that automatically gets complex shapes and speeds up the convergence is used. A numberof medical ultrasound images are applied to verify the performance of the proposed model and compared with the traditional active contour model of local Gaussian distribution fitting with regularization item and the recently proposed fast segmentation algorithm of multiscale and shape constrainedlocalized C-V. The experimental results reveal that this algorithm overcomes the fast segmentation problem of the medical ultrasound images, and the calculation efficiency is 49.4% higher than the one of multiscale and shape constrained localized C-V model.