Polar Coordinates Texture Analysis Methods for Digitized Intravascular Images

Abstract

The image texture of ultrasound images correlates highly to the histological composition of the corresponding insonified tissue. Therefore, tissue identification can also be performed by analyzing textural features. Many methods of digital image texture analysis are based on measuring the relationship between pixels in a given environment. This environment is described in terms of pixel deviation in both horizontal and vertical directions. Intravascular images show different tissues such as thrombi, soft or calcified plaques, as well as the various wall layers arranged in concentric rings around the ultrasound catheter. Due to the circular geometry of blood vessels, regions of interest are better defined as circular portions of sectors than as quadratic windows as usually considered in image processing. Additionally, spatial resolution decreases with growing radii such that the normally computed texture spatial features, i.e. gray level run length or co-occurrence properties, should not be equally treated in all image regions. The texture appears to be finer with smaller radii, although for a histologically uniform region, tissue structure may not be different. Texture analysis methods were therefore computed in polar coordinates, as presented in the first section.