Meeting Banner
Abstract #2993

High Resolution 3D Intracranial ASL Angiography Using Automatically Tuned Compressed Sensing

Huimin Wu1, Walter F. Block1, 2, Patrick Turski3, Charles A. Mistretta1, Kevin M. Johnson1

1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States; 2Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States; 3Radiology, University of Wisconsin-Madison, Madison, WI, United States

Fully 3D compressed sensing was implemented with iterative soft thresholding (IST) and used for reconstruction of 3D intracranial angiograms obtained with PCASL-VIPR. A data-driven threshold selection method based on SURE (Steins Unbiased Risk Estimator) was utilized at each iteration to adaptively calculate the threshold. Standard non-iterative reconstruction, auto-tuned CS and manually tuned CS were compared on a static PCASL-VIPR dataset with an isotropic resolution of 0.43 mm. CS was able to greatly improve the SNR and vessel visualization with acceleration as high as 20X without manual parameter tuning.

Keywords

accelerated acceleration achieve achieving acquisition amenable analyzes aneurysms angiography apparent applying arterial artifacts assessment assistance assoc auto axial biomedical block challenges challenging channel coefficients coil combine combined comm comparing complexities compressed computational considerable consistency control coverage created currently defined denotes descent diagnosis driven duration dyadic encoding encourages engineering enhanced error estimator exist feasibility flight flow free fully function furthermore gradient grant gratefully head highly human improvements include individual inferior instruments insufficient interfere isotropic iteration iterative iteratively length limited manually math matrix medical minimization minimizing near noise optimal optimally optimized parallel passage physics preliminary preserving principle pure quality radial radiology rapid recast reconstructed reduced relatively represents requires resolution respectively risk routine sampling saturation scanner selected selection sense sensing sensitivity sets soft solution space sparsity spatial spin square stein step still studies substantial subtracted sure tagging tends term terms thresh threshold thresholding thresholds transform transforms tuned tuning unbiased vessel visualization wavelet whole window