Artery-Vein Detection in Very Noisy TOF Angiographic Volumes using Dynamic Feedback Scale-space Ellipsoidal Filtering

J.S. Suri (USA), K. Liu (PRC), L. Kasuboski (USA), S. Singh (UK), and S. Laxminarayan (USA)


Magnetic Resonance Angiography (MRA), Black Blood Angiography (BBA), Pre-filtering, Median, Directional, Scale-Space, Ellipsoidal, Vessels, Arteries, Veins, Quan tification, Stenosis, MIP


Pre-filtering is a critical step in 3-D segmentation of the blood vessel and its display. This paper presents a feedback scale-space approach for filtering the white blood angio graphic volumes. The raw MR angiographic volume is first converted to isotropic volume followed by 3-D higher or der separable Gaussian derivative convolution with known scales to generate edge volume. The edge volume is then run by the directional processor at each voxel where the eigenvalues of the 3-D ellipsoid are computed. The ves sel score per voxel is then estimated based on these three eigenvalues which suppress the non-vasculature and back ground structures yielding the filtered volume. The fil tered volume is then scale-space thresholded using a dy namic threshold which is computed using a combination of Bayesian threshold and a scale-dependent decay func tion. For complete capture of the vessels in the volume, the scales are made to increase from Ñ Ò to Ñ Ü and then optimized. For each scale, a new threshold is com puted thereby making the system design dynamic. The in creasing scales to capture thick vessels uses this dynamic threshold each time the scale changes, hence is dynamic to the changing input scale. We demonstrate this system for cartoids in a very noisy data set on white blood angiogra phy volumes. We qualitatively and quantitatively measure the performance of the system by computing the MIPs and SNR/CNRs. We run our system over more than 20 patient studies.

Important Links:

Go Back