We implemented two advanced reconstruction methods for highly accelerated subtractive NCE-MRA, which can exploit the sparsity of subtracted angiograms. One method is based on k-space subtraction of complex raw data with phase and intensity correction (KSPIC). Another method is to reconstruct bright- and dark-blood data with an additional magnitude subtraction term in the cost function to exploit the sparsity. The performance of the two methods was evaluated in both retrospective and prospective accelerated datasets using quantitative metrics and qualitative scoring. Compared with conventional methods, they both showed improved image reconstruction quality, while KSPIC had the best performance.