Automatic Characterization Study Of Atherothrombotic Plaques Based On Intravascular Ultrasound Images

Cardiovascular disease has become the leading cause of death in the world,more and more people are dead from coronary heart disease(CHD).Atherosclerosis is a kind of proliferative inflammation that affects arterial blood vessel wall,becoming the main cause of coronary heart disease.CHD is the latent state that causes the fatal disease of cardiovascular and cerebrovascular diseases.By placing tiny ultrasound transducers on the top of the catheter into the blood vessels,intravascular ultrasound(IVUS)captures the images of the endovascular and vascular walls during the catheter retreats.At present,IVUS provides a clinically visual tools for the blood vessels.Using IVUS,the information of the composition and morphology of atherosclerotic plaque could be obtained to evaluate vulnerability of the plaques and predict the risk of CHD.The IVUS images are constructed with the ultrasonic echoes of different tissues in the vascular wall showing its structure in cross-sectional view.The traditional method of identifying the plaque tissues in the IVUS inmages is based on the doctors’experiences.However,this manual identification method is subjective and cumbersome due to the large amount of IVUS images.Therefore,an efficient automatic identification method of the plaque tissues in the IVUS images is needed.This study proposed an automatic method to recognize and classify the region of interest(ROI)of plaque tissues in IVUS images.The results show the potentials of the proposed method to evaluate coronary atherosclerotic lesions accuracy atherosclerotic lesions accurately based on the plaque features.It is well known that the IVUS images are rich in gray-scale information and texture information.Therefore,based on a variety of texture features extracted from the IVUS images,with use of proper classifiers,it is convenient to identify the different types of the plaque tissues.According to the characteristics of IVUS images,this study presents an effective algorithm of identification of the plaques in the IVUS images.According to the IVUS imaging principles,the calcified plaques have obvious features,that is,acoustic shadowing is often formed in the posterior region of calcified plaques with hyperechogenicity.There are artifacts generated during the process of IVUS imaging.Using the characterization acoustic shadowing,the insight exploration of identifying the calcified plaque area was performed in this study.An automatic identification method of atherothrombotic plaques in IVUS images based on the plaque characterization was proposed.In this study,two databases(i.e.the database of IVUS plaques and the database of the calcified plaques)were established by the collecting IVUS data,selecting before marking the ROI of the IVUS images.(1)The plaques samples database:The IVUS images from 50 patients with cardiovascular disease were collected.Finally,143 frames of cardiovascular disease from ten patients were selected to establish the dataset,which included 207 plaque samples contain,i.e.91 fibrotic plaque tissues,66 lipid plaque tissues and 50 calcified plaque tissues.(2)The calcified plaque samples database:Calcified plaque database was selected from 50 patients with cardiovascular disease.According to the identification of the doctors,100 frames of IVUS images with the calcified plaque were selected to establish the database.After establishing the datasets,the identification of atherosclerotic plaques were performed.1.The aim of this study is to apply a machine learning method to identify the different types of plaque tissues in intravascular ultrasound(IVUS)grayscale images.In this study,207 plaque samples in the IVUS images from 10 patients with cardiovascular disease in the hospital were analyzed.Firstly,a sliding patch was selected and its center pixel traversed in the plaque area.The values of the mean and entropy were calculated.Ten features of the patch along 4 directions were respectively obtained using co-occurrence Matrix method.Secondly,more texture features of the region of plaque in the IVUS images were obtained using Gabor filter and Local Binary Pattern(LBP)methods.Finally,the classifiers of Liblinear,random forest and Harmonic to Minimum-Generalized LVQ(H2M-GLVQ)were used to classify these pixels in the plaque tissues based on the features obtained by reducing dimension by principal component analysis(PCA).The manual characterization by an experienced physician was considered as the gold standard.Results of the proposed automatic characterization method show the general identification rate of classifiers of random forest and H2M-GLVQ is over 80%.In comparison with the other two classifiers,the identification rate of random forest was relatively higher,i.e.89.04%、80.23%and 73.77%respectively for fibrotic,lipidic and calcified plaque tissues.2.The aim of this study is to perform an insight investigation on the efficient method to identify the calcified plaque in intravascular ultrasound(IVUS)grayscale images based on the specific characteristic of acoustic shadowing.Firstly,the image was denoised by bilateral filtering.The "circle" image is transformed to the"rectangle" image by using polar inverse transformation.Secondly,the preprocessed image was clustered into 7 areas by K-Means method and then smoothed by the median filtering.Finally,the ratio of the average gray value of ROI to the posterior region of the ROI was compared with the preset threshold to judge whether the ROI is calcified plaque or not.Then the image was transformed back into 512×512"circle" image.The results show that the accuracy of classification of calcified plaque improved reaching up to 78.67%in comparison with the former method combining texture information and classifiers.In conclusion,this study applied the characterization methods to identify the plaques in blood vessels.The methods have potentials to classify the calcified,lipid and calcified plaque tissues and to evaluate the stability of the plaques.