Moment-based texture segmentation of luminal contour in intravascular ultrasound images

Purpose. A system for luminal contour segmentation in intravascular ultrasound images is proposed. Methods. Moment-based texture features are used for clustering of the pixels in the input image. After the clustering, morphological smoothing and a boundary detection process are applied and the final image is obtained. Results. The proposed method was applied to 15 images from different patients, and a correlation coefficient of 0.86 was obtained between the areas of lumen automatically and manually defined. Conclusion. Moment-based texture features together with the radial feature are powerful tools for identification of the lumen region in intravascular ultrasound images. Morphological filtering was useful for improving the segmentation results.