Segmentation of the luminal border in intravascular ultrasound B-mode images using a probabilistic approach

► Lumen parametrization using Fourier series. ► Lumen contour is deformed by the minimization of a probabilistic cost function. ► The likelihood of the pixels in the images to belong to lumen are obtained using probability estimates provided by a Support Vector Machine classifier. ► We introduce an optimization strategy which improve convergence. ► The proposed method is capable of segmenting IVUS images from different frequencies without the need of parameter tuning. ► The proposed method is robust with respect to changes of the B-mode reconstruction parameters.