Automated quantification of arterial stenosis on CE-MRA by using a deformable vascular tubular model

Methods Contrast-enhanced MRA from 21 patients were included. MR images were acquired using a 1.5T MRI scanner with a spoiled 3D FLASH acquisition and a 4x2 circularly polarized phased-array neck coil. Four consecutive 3D images were acquired starting at approximately 3s after the administration of 0.1 mmoL/kg gadolinium. Subtraction images were generated to improve vessel-to-background contrast. To demonstrate the methods’ robustness against image noise, nine subtraction images were included. Curved multiplanar reformatted image slices were generated perpendicular to the vessel direction and an expert drew luminal contours on these slices. A fullwidth half-maximum criterion was applied to maintain the contour consistency, particularly for low vessel-tolumen contrast area. The user defined the artery section of interest by placing proximal and distal points. Subsequently, an