Meeting Banner
Abstract #4481

Application of PLS-DA Method in Separation of Venous and Arterial Phase in 4D CE-MRA Data

Siamak Salari Sharif1, Ewart Mark Haacke1, Joseph J. Hewett2, Michael Arata2

1MR Innovations Inc., Detroit, MI, United States; 2Synergy Health Concepts, Newport Beach, CA, United States

In contrast-enhanced MR-angiography (CE-MRA), often, it is desired to extract a single phase; venous or arterial, for further processing or visualization. This task can be very complicated especially in patients with venous abnormalities as the flow is accompanied with various delays or is obstructed, leading to inability to identify a clear arterial phase in the image for subtraction. Therefore, we developed an algorithm based on partial least-squares discriminant analysis to isolate the venous/arterial phase in 4D CE-MRA. The advantage of our method is its ability to handle highly correlated MRI datasets, as well as capturing the variability in contrast enhancement.

Keywords

ability abnormalities accompanied acquiring addition advantage agent amounts analyze angiography application applied arterial arteries assigning background baker beach belonging better choose class classified clear coefficients complicated concatenated concepts consecutive consents construct contrast correlated correlation create dataset datasets dedicated delays desired determine developed directly discriminant effectively effectiveness elsewhere enhanced especially every fields finally flow form geometry handle health highlighting highly individual injected injection injector innovations input insensitive intensities intensity interleaved investigator involved isolate isotropic jersey latent least likelihood locations maps mark matrix measuring next obstructed often onto orthogonal partial patients place predicted prediction pressure process processing project projections protocol rank reconstructing refold refolded regression removing resolved rest scanner scanning selected selects separate separately separation simple smoothed squares stage step steps stochastic structure subjects subtraction successfully synergy task temporal took train trained training trajectories trio twist unfold unfolded unfolding useful user variable various vector vectors veins venous vessels visualization visualize visualizing volume volumes