 |
|
Intravascular Ultrasound Image Segmentation
Intravascular Ultrasound (IVUS) is a catheter-based technique, which provides real-time high-resolution images allowing precise tomographic assessment of lumen area, plaque size, and composition of a coronary segment, and therefore provides new insights into the diagnosis of and therapy for coronary disease.
In IVUS images, the lumen is typically a dark echo-free area adjacent to the imaging catheter, and the coronary artery vessel wall mainly appears as three layers: intima, media, and adventitia (Fig.1). In clinical research, the two inner layers are of principal concern.
Figure 1. An IVUS image
Segmentation of IVUS images to isolate the intima-media and lumen provides important information about the degree of vessel obstruction as well as the shape and size of plaques. However, manual segmentation is time-consuming and susceptible to intra- and inter-observer variance.
Based on the fact that the interaction of ultrasound with coronary arterial structures gives rise to typical image patterns that may be used to identify the lumen and arterial wall in ultrasound images, we have developed an automatic method to segment the plaque or intima-media area of coronary arteries in IVUS images based on a deformable model, which integrates both edge and region information. A “time-delayed” discrete dynamic programming approach is adopted to implement this method. The key characteristic of our method compared to other methods used in IVUS image segmentation is that it uses contrast information for edge detection, rather than the gray level gradient in the images.
Some segmentation results obtained by our method are presented below. Since several studies have reported good correlation of expert-determined measurements with histology, we have validated our method by comparing the boundaries obtained by automatic segmentation with corresponding hand-traced ones (Fig. 2). The lower images, which were scanned after the contours had been drawn by the expert, have slight deformations compared to the computer-generated results. However, from the local image textures, it can be seen that the contours generated by our algorithm are very close to the ones traced by the experts.
Figure 2. Segmentation results: upper ones are by our algorithm; lower ones are manually traced by an expert.
| Back to Research |