Automatic Segmentation of Soft Plaque by Modeling the Partial Volume Problem in the Coronary Artery

Automatic segmentation and quantification of stenosis is an important task in assessing coronary artery disease, especially when the investigation of the disease progress is considered. The reproducibility and robustness of the segmentation algorithm against partial volume effect and noise is critical for an accurate quantification. A major issue in the quantification of the stenosis is to segment the soft plaque in the blood vessel. While there are several approaches for segmentation of the volume of the blood vessel and soft plaque in the literature, the main drawback of these approaches is making a deterministic decision in terms of assigning a particular voxel to only one type of tissue (such as blood vessel, soft plaque or surrounding area). However in reality, because of the partial volume effect, a voxel may contain more than one tissue type. In particular, using deterministic methods for quantification of the small objects such as thin blood vessels or soft plaque may lead to inaccurate results and higher inter and intra-scan variability. In this paper, an approach is proposed to tackle the partial volume effect problem using an adaptive fuzzy algorithm incorporating a Markov random field model. The presented method segments the blood vessel, soft plaque and surrounding tissue areas more accurately. The algorithm is applied to several datasets and the outcomes have been judged visually by a qualified radiologist. The proposed algorithm has the potential to be applied for the accurate quantification of the degree of stenosis.