Robust Anatomy Detection Using Tracking Algorithms in 4D Echocardiography

Abstract

Cardiac diseases are considered to be the most lethal class of diseases, impacting approximately 1 in 3 individuals in Western countries. Currently, the vast majority of surgical procedures targeting these afflictions involve percutaneous, non-invasive methods, which rely heavily on various imaging techniques, along with advanced image guidance through computer vision. This article aims to present an assessment of how various detection and tracking techniques such as Marginal Space Learning, Optical Flow, and Belief Propagation can be combined to identify anatomical structure locations in 4D echocardiography, at high speed (34 ms) and with great accuracy (2.5 ± 1.2 mm mean error).