Coronary Atheroclerotic Plaque Characterization Based On Intravascular Optical Coherence Tomography Imaging

Coronary artery disease(CAD)is the leading cause of death in the world and atherosclerotic plaques are the main culprit to blame.Identifying the plaque type is of great significance for the prevention of coronary syndrome.Intravascular optical coherence tomography(IVOCT)is rapidly becoming an effective method for diagnosis of CAD due to the ultra-high image resolution(10-20μm).It performs cross-sectional imaging of coronary arteries by detecting the backscattered light from the tissue.Currently,the IVOCT image analysis in clinic has been limited to manual procedure,which is not only time consuming,but also inaccurate due to inter and intra observer variability.The aim of this project is to develop an algorithm for automatically identifying the plaque type in IVOCT images.Main steps of IVOCT image analysis are as follows.Firstly,image preprocessing was conducted to extract the lumen border.This includes guide-wire artifacts removal(via dynamic programming method)and catheter removal(by Otsu’ method and some morphological operation)etc.Accurate lumen information is the only effective metric for evaluating the vascular stenosis.Our results show that the Dice coefficient between the automatically segmented lumen and the manually segmented results is 0.9413.After that,7 texture features were extracted for different plaques by using gray level co-occurrence matrix(GLCM).They were contrast,correlation,energy,homogeneity,cluster shade,entropy and information measures of correlation.We also investigated the correlation between the plaque class and texture features under different GLCM parameters.After feature selection,the feature dimension was reduced to 378.At last,random forest method was used to model the extracted texture feature.For a new unclassified pixel,we can use the constructed model to do the classification.Overall pixel-wise classification accuracy is 78%,and per-class accuracy is 84%,77%,72%for fibrous,calcified and lipid-rich plaque respectively.Hereto,an algorithm has been developed for automatic atherosclerotic plaque characterization according to texture features.It could be useful for clinical IVOCT image analysis.