Optical coherence tomography (OCT) imaging has been widely employed in assessing cardiovascular disease.
Atherosclerosis is one of the major cause cardio vascular diseases. However visual detection of atherosclerotic plaque
from OCT images is often limited and further complicated by high frame rates. We developed a texture based
segmentation method to automatically detect plaque and non plaque regions from OCT images. To verify our results we
compared them to photographs of the vascular tissue with atherosclerotic plaque that we used to generate the OCT
images. Our results show a close match with photographs of vascular tissue with atherosclerotic plaque. Our texture
based segmentation method for plaque detection could be potentially used in clinical cardiovascular OCT imaging for
Atherosclerotic coronary artery disease continues to be one of the major causes of mortality. Prevention, diagnosis and treatment of atherosclerotic coronary artery disease are dependent on the detection of high risk atherosclerotic plaque.
As age is one of the most important risk factors, atherosclerosis worsens steadily with increasing age. Automatic characterization of atherosclerotic plaque using the optical coherence tomography (OCT) images provides a powerful tool to classify patients with high risk plaque.
In this study we develop an automatic classifier to detect atherosclerotic plaque in young and old Watanabe heritable hyperlipidemic (WHHL) rabbits, using OCT images without reliance on visual inspection. Our classifier based on texture analysis technique may provide an efficient tool for detecting invisible changes in tissue structure.
We extracted a set of 22 statistical textural features for each image using the spatial gray level dependence matrix (SGLDM) method. An optimal scalar feature selection process was carried to select the best discriminating features that employ the Fisher discriminant ratio (FDR) criterion, and cross correlation measure between the pairs of features. Using these optimal features, we formed a combination of 5 best classification features using an exhaustive search method. A combined feature set was finally employed for the classification of plaque. We obtained correct classification rate and validation of 76.67% and 75% respectively.