Classification of Intravascular Ultrasound Signal by Kernel Density Estimation and Bayes Theorem for Identification of Coronary Plaque Tissue

We propose a method for tissue characterization of coronary plaque using IntraVascular UltraSound (IVUS) data. In the proposed algorithm, the probability density function of each plaque is calculated by the kernel density estimation and neural gas. The probability distributions vary continuously and overlap one another in the feature space. The class labels of tissue are determined by using Bayes estimation. This algorithm enables to give more delicate information to a medical doctor, which matches with the real situation of plaque composition. In this paper, the proposed algorithm is presented brief and its effectiveness is confirmed by applying it to the real IVUS data.