Coronary Plaque Boundary Detection in Intravascular Ultrasound Image by Using Hybrid Modified Level Set Method and Fuzzy Inference Syaiful Anam a,b , Eiji Uchino a,c and Noriaki Suetake a a Graduate School of Science and Engineering, Yamaguchi University 1677-1 Yoshida, Yamaguchi 753-8512, Japan {r501wa, uchino, nsuetake}@yamaguchi-u.ac.jp b Department of Mathematics, University of Brawijaya Jl. Veteran Malang 65145, Indonesia c Fuzzy Logic Systems Institute 680-41 Kawazu, Iizuka 820-0067, Japan ABSTRACT This paper describes a boundary detection of coronary plaque by a hybrid of a modified level set method and a fuzzy inference. The standard level set method to detect an image boundary commonly uses an image gradient for calculating a speed function. But the speed function of the level set cannot work well within an intravascular ultrasound (IVUS) image, which is the target of this paper. Therefore, we proposed a method for coronary boundary detection in IVUS image by using a modified level set method. In that method, the image gradient in the speed function is substituted by the weighted image separability. However, some regions of the IVUS image often becomes shadowed and then contains no texture information, due to the presence of the guide wire. Thus the modified level set method fails to detect the plaque boundary in those regions. To solve this problem, we further propose in this paper, a hybrid of the modified level set method and the T-S fuzzy model. The present method has been more successful in the accuracy of plaque boundary detection. KEYWORDS Coronary plaque, boundary detection, intravascular ultrasound image, modified level set method, T-S fuzzy model. 1 INTRODUCTION Acute coronary syndromes (ACS) is one of the leading reasons for hospitalizations in the world. It is caused by a rupture of vulnerable plaque which is built up inside the coronary arteries causing heart attacks. One of the medical imaging methods to diagnose ACS is the intravascular ultrasound (IVUS) method [1]. IVUS images are used for the inner and outer coronary plaque boundary detection to evaluate the quantitative assessment of the coronary plaque compositions. Detecting coronary plaque boundaries is a very difficult task for medical doctors. In order to reduce the workload of the medical doctors, an automatic coronary plaque boundary detection method with high accuracy has become a necessity. Previous works [2] and [3] were proposed to solve this problem. The fuzzy inference model employed in the previous works is Takagi-Sugeno (T-S) fuzzy model [4]. Level set methods have been widely used in image segmentation and have several advantages over other segmentation methods such as the snake method, region growth and thresholding. The standard level set method commonly uses a gradient for calculating a speed function, but the speed function of the level set cannot work well within an IVUS image. Therefore, we proposed a method for coronary plaque boundary detecion by using the weighted image separability method and the level set method [5]. In [5], the image gradient in the speed function is substituted by the weighted image separability. However it failed to detect the plaque boundary for several regions of the image as shown in Fig. 1(b). The plaque boundary detection in IVUS image is very difficult because a region of the IVUS image often becomes shadowed and then contains no texture information due to the presence of the guide wire as shown in Fig 1. Thus the method [5] fails to detect the plaque boundary. ISBN: 978-0-9891305-2-3 ©2013 SDIWC 69