Investigation and Classification of Cardiac MRI Images using Texture Analysis

Abstract

In order to quantify the myocardial viability after infarction, recent researches have developed new cardiac MRI protocol using particle contrast agents which highlight the different areas after infarction: i.e. the healthy area, the necrosed area and the two ischemic areas. Actually the researches are carried out on rodents micro cardiac MRI for which the obtained images are very noisy and present a very low contrast. In this paper, we analyse the potential of texture analysis to classify each pixel of the myocardium, on rodents micro cardiac MRI, in one of the three areas. In a single image, the region of interest is selected, i.e. ROI and find linear separatblity coefficient and apply Linear Discriminant Analysis on selected region of interest which gives exact classified area in the image. Then we performed Texture analysis is used to separate the unwanted region from the desired region, allowing the desired region to be found. Texture refers to the representation of an image or the concepts that surround it. Texture analysis is used extensively in object recognition, surface defect detection, pattern recognition, medical image analysis, and other computer vision applications. Three types of ROI in a single image with different pixels are considered for Texture analysis. An image is divided in to two distinct areas each has different texture. Now texture image is segmented through which the image boundaries are defined and areas are compared. If the boundaries and other characteristics are different the range of the boundaries can be defined. The technique of "Texture Segmentation" divides an image into distinct areas, each with a different texture. Texture segmentation is the process of defining the boundaries of various textures. In other words, texture segmentation compares the features of the boundaries and areas and determines the boundary range if their texture characteristics are sufficiently different. The texture analysis was then applied to a very small size images does not seem to be a way for further searches. But we can expect better results on human cardiac MRI for which resolution and contrast would be much better.