Automatic segmentation of intravascular ultrasound images based on temporal texture analysis


引用 23|浏览7
In our study we developed a novel automatic algorithm for the analysis and delineation of lumen and external elastic membrane (EEM) boundaries using both temporal and spatial variation of IVUS data. The pre-processing steps involve the construction of Laplacian gradient image from neighboring images, and the use of discrete wavelet frame decompositions for texture computation. A smooth Lumen and EEM contour is predicted by applying radial basis functions on contour initialization. This algorithm is evaluated on large data set of multi-patient 2293 IVUS images and pitted against the manually segmented contours by medical experts. It is observed that this algorithm reliably performs contour prediction with clinically appreciated limits of average prediction error equaling 0.1254 mm and 0.0762 mm for Lumen and EEM respectively. Furthermore a custom Lumen detection algorithm for stented images is proposed and tested with average prediction error of 0.048 mm.
biomechanics,biomedical ultrasonics,biomembranes,discrete wavelet transforms,elasticity,image segmentation,image texture,medical image processing,eem boundaries,eem contour,laplacian gradient image,lumen detection algorithm,contour initialization,discrete wavelet frame decompositions,external elastic membrane boundaries,image preprocessing,intravascular ultrasound image segmentation,multipatient ivus images,radial basis functions,smooth lumen contour,stented images,temporal texture analysis,texture computation,image recognition,biomedical imaging
AI 理解论文
Chat Paper