IVUS Tissue Characterization of Coronary Plaque by Classification Restricted Boltzmann Machine

The tissue characterization of coronary plaque is an important task to assess the atherosclerotic process and the potential risks of their ruptures on patient. Thanks to intravascular ultrasound (IVUS) medical imaging technique, the reflected ultrasound signals from tissues are acquired, then be used to visualize inside the artery by the computer-assisted equipment. Often, the characterization of tissues is based on the analysis of their responding echo intensity. However, the domination of various factors and the data robustness are the realistic challenges of IVUS classification problems. The quality of the visualization totally depends on the proposed classifier of descriptive features along with its algorithm. In this study, our objective is to characterize IVUS tissues by using classification restricted Boltzmann machine (ClassRBM). We propose to binarize feature patterns extracted from time domain signals for the input of ClassRBM. The results show a better evaluation compared to the conventional integrated backscatter IVUS method (IB-IVUS) for the same task.