A new method for single-step robust post-processing of flow Color Doppler M-mode images using support vector machines

Intra-cardiac pressure gradients (ICPG) are usually estimated by post-processing of flow Color Doppler M-mode images (CDMMI) by using a sequence of processing steps. We propose a novel image processing method which gives a single-step approximation of the ICPG image, based on a simple, yet specifically developed, Support Vector Machine (SVM) algorithm. Our method only requires the SVM estimation of the blood velocity from the CDMMI. Given that ICPG images are obtained by deterministic operators (Euler’s momentum equation) on the blood velocity, the ICPG estimation is a simple model that consists of the same coefficients and the operator applied to the Mercer’s kernel. A diverse-width Mercer’s kernel is proposed, as an alternative to conventional Radial Basis Function kernel. Simulations on a synthetic model and approximations of a real example image, trained with up to 10% of the pixels, show the possibilities of this new single-step post-processing method.