Jiang Du1, Afshin Karami1, Yijing
Wu2, Frank Korosec2, Thomas Grist2, Charles
Mistretta2
1Radiology, University of California,
San Diego, CA, United States; 2Medical Physics and Radiology,
University of Wisconsin, Madison, WI, United States
Time-resolved
CE-MRA provides contrast dynamics in the vasculature, which can be further
used to separate arteries from veins. However, most of the segmentation
algorithms require operator intervention. Furthermore, the contrast dynamics
pattern may vary significantly within a large coronal imaging FOV due to
delayed or asymmetric filling, or slow blood flow in the tortuous vessels.
Correlation with single arterial and/or venous reference curves may result in
misclassification. Here we present a fully automated region-specific
segmentation algorithm for effective separation of arteries from veins based
on cross correlation and pooled covariance matrix analysis.