Automated contour detection for high-frequency intravascular ultrasound imaging: A technique with blood noise reduction for edge enhancement

Automated edge detection may standardize measurements among observers, prov iding for rapid assessment of intravascular ultrasound (IVUS) images. Howev er, with high frequency images, enhanced blood signals make it difficult to define and trace the lumen borders. Accordingly, we evaluated a fully auto mated contour analysis facilitated with a blood noise reduction algorithm ( BNR) for 40-MHz IVUS images in human coronary arteries of 27 patients. This algorithm is based on the principle that blood echo speckles have higher t emporal and spatial variations than the arterial wall, A total of 193 paire d lumen areas and 78 external elastic membrane (EEM) areas were measured an d compared. Automated measurements showed good agreement with manual tracin gs for lumen and EEM area, with high correlation coefficients (0.945 and 0. 950, respectively) and small variability (0.4 +/- 14.4% and 0.6 +/- 9.7%, r espectively). This preliminary finding suggests that automated contour dete ction facilitated with BNR appeared to be a feasible and reliable technique for area measurements in 40-MHz IVUS imaging. (C) 2000 World Federation fo r Ultrasound in Medicine & Biology.