Ultrasound Information Analysis In Atherosclerotic Hardened For Distinguishing,

Atherosclerotic disease is common in elderly people.In presence of such disease,the artery may be subjected to morphological changes that can perturb the normal blood circulation.In asymptomatic subjects it may represent a risk index for some severe pathologies like acute stroke or cardiac infarction.Early diagnosis and assessment of atherosclerosis is of paramount importance.The ultrasonic scan is widely used in the clinical diagnosis of carotid atherosclerosis.However,such ultrasonic scan is a time-consuming procedure,mainly depended on the subjective operator assessment,which inevitably leads to large work load for operators and inter intra-observer variability. Therefore,it's necessary to establish a computer aided automatic recognition system with the high performance.The dissertation focuses on extracting and assessing parameters of carotid artery by using different ultrasonic techniques,to classify carotid atherosclerosis automatically.The dissertation is organized as follows:1.The intima-media thickness(IMT) and lumen diameter(LD) of common carotid artery serve as early markers of carotid atherosclerosis.Increasing IMT is demonstrated to have strong correlation with the presence of atherosclerosis elsewhere in the body.Hence, given the fact that each pixel of ultrasound images may be regarded as a node in a directed graph,a method based on heuristic graph searching algorithm(A* algorithm) is proposed to detect the near-wall intima,far-wall intima and media.IMT and LD are calculated according to extracted boundaries.2.Carotid atherosclerosis will result in the change of various hemodynamic parameters. The maximum flow velocity curve,carrying much information of blood flow,is of great clinical significance.Here,the maximum flow velocity curve is firstly extracted from the Doppler ultrasound spectrum.The above-mentioned heuristic A* algorithm is used to extract such curve from internal carotid artery and vertebral artery,on which several spectrogram parameters are calculated.Then,together with IMT and LD,six large Mahalanobis distance of blood features are chosen to be used in the computer aided recognition system.3.The Adaboost algorithm is employed in the recognition system.Adaboost is a method to find a highly accurate classifier on the training set by combing weak hypotheses. Here,the BP neural network serves as weak hypotheses.Seven features including IMT and spectrogram parameters are selected as the feature vector that best describe carotid artery characteristics.After several experiments with different conditions,it is found that the Adaboost-BP network improves the performance of individual BP network.IMT scores incorporating data from spectrogram parameters are well correlated with the extent and severity of carotid atherosclerosis,and better than the individual IMT scores or individual spectrogram parameters.Finally,a computer aided automatic recognition system is constructed based on above three parts.Experiments on 35 pathologically proven cases including 10 normal carotid arteries and 25 abnormal ones show that the proposed system yields an accuracy of 94.12%, a sensitivity of 91.67%,a specificity of 100%,a positive predictive value of 100%and a negative predictive value of 83.33%.Therefore,it is concluded that the proposed system performs well in the ultrasonic classification of carotid atherosclerosis.Hopefully,the system may act as an effective assist diagnosis tool in the clinical application environment.