The Research On Automatic Plaque Segmentation Algorithm In Coronary OCT Images

With the development of society,people’s quality of life continues to improve,cardiovascular disease has become a serious threat to the quality of human life.Cardiovascular disease is a common preventable cause of death.At present,there are many kinds of medical imaging diagnostic techniques for cardiovascular diseases.In order to identify the components of the lesions more clearly,experts can use the intravascular optical coherence tomography(OCT)to detect and characterize the lesions.However,the focus diagnosis of coronary OCT image still needs expert manual annotation,which not only results in segmentation error due to expert individual differences,but also takes time and effort.Therefore,the realization of artificial intelligence automatic segmentation of coronary OCT images is of great significance for the clinical diagnosis of cardiovascular disease.In this paper,the artificial intelligence and deep learning method are used to realize the automatic segmentation of three kinds of characteristic plaque in coronary OCT image.The main contents of this paper are as follows:(1)This paper proposes a method for plaque segmentation of coronary OCT images based on improved neutrosophy theory.According to the characteristics that neutrosophy theory can classify pixels of differet regions in the image.In this paper,the s-membership function is used to change the neutrosophy conversion formula to accurately classify the pixels.The accuracy of pixel classification is improved,and then the accuracy of image segmentation is improved.In this paper,the method can use the transformation process of the neutrosophy algorithm to automatically segment the fiber plaque and lipid area,and combine the two transformed domain images to segment the calcified plaque accurately using the region growth algorithm.(2)This paper proposes a multi-region automatic segmentation algorithm for coronary OCT images using a hybrid model of U-Net model and membership-neutrosophic algorithm.Based on the small data set of coronary OCT images,this paper simplifies the U-Net network model,uses the template manually labeled by the doctor as a label,trains threefeature plaques separately,and finally inputs the rough segmentation results into the membership-neutrosophic algorithm model for refinement segmentation.Among them,the membership-neutrosophic model is a custom image conversion method,which changes the method of converting to the neutrosophic domain in the traditional neutroscopy algorithm,making the pixel classification more accurate.This method can automatically and accurately segment the three characteristic plaques in coronary OCT images.The segmentation effect and accuracy are verified from many aspects,and the effectiveness of the hybrid model is proved.