Abstract

The manual inspection of intravascular ultrasound (IVUS) images to detect clinically relevant patterns is a difficult and laborious task performed routinely by physicians. In this chapter, we have investigated the use of deep convolutional nets for the quick selection of IVUS frames containing calcified plaque, a pattern whose analysis plays a vital role in the diagnosis of atherosclerosis. Our networks are designed to detect an entire segment of an IVUS sequence as clinically relevant for the pattern of interest. A sequence-based postprocessing is applied to the network outputs exploiting prior knowledge on the temporal behavior of the ground-truth signals. Our preliminary experiments on a dataset acquired from 80 patients and annotated by one specialist showed that deep convolutional architectures improve on a shallow classifier by a significant margin.