Intravascular optical coherence tomography image segmentation based on Gaussian mixture model and adaptive fourth-order PDE

The accuracy of the fibrotic plaque segmentation is vital in identifying the coronary artery stenosis. In this paper, we address an automated approach (APDE-GMM) for separating the fibrotic plaque area of intravascular optical coherence tomography (IV-OCT) images. Under this approach, an objective function consisting of a new energy functional with Rayleigh distribution and the negative log-likelihood function of Gaussian mixture model (GMM) is developed. Also, the study presents an adaptive diffusivity function where the gradient threshold can be associated to suppress the effect of speckle noise. The parameter estimation is carried out by the expectation–maximization technology. In addition, this paper derives a fourth-order partial differential equation (PDE) via Euler–Lagrange equation to obtain the optimal solutions. It has been compared to other segmentation approaches on synthetic and clinical IV-OCT images. The results demonstrate that APDE-GMM segmentates more accurately.