Kedar Khare1, Christopher J. Hardy1, Kevin F. King2, Patrick A. Turski3, Luca Marinelli1
1GE Global Research Center, Niskayuna, NY, United States; 2GE Healthcare, Waukesha, WI, United States; 3School of Medicine & Public Health, University of Wisconsin, Madison, WI, United States
We present a robust method for compressed-sensing reconstruction using a data-driven, iterative soft-thresholding (ST) framework with no tuning of free parameters. The algorithm combines a Nesterov-type optimal gradient scheme for iterative update with adaptive wavelet denoising methods. Vascular 3D phase-contrast scans on healthy volunteers are used to show that image quality is comparable to that of empirically tuned, nonlinear conjugate-gradient (NLCG) reconstruction. Statistical analysis of image quality scores for five datasets indicates that the ST approach improves the robustness of the reconstruction and image quality as compared to NLCG with a single set of tuning parameters for all scans.