Super Resolution in MRI: How far can we go?

Super Resolution (SR) in MRI has been described as a technique that allows an increase of resolution without a loss in the Signal to Noise Ratio (SNR), essentially providing "free" resolution. This study explores this claim and investigates the limits to SR in the presence of noise. We show that the amplification of noise due to the SR algorithm can outweigh the signal benefit, leading to SNR r eductions that could counter the motivation for using the SR approach. Thresholds for successful SR approaches are determined through simulation, which are lower than used previously in the literature. We also assess the initia l resolving power of different slice profile shapes, without any further post-processing, and show that for rectangular slice profiles, it can be far higher than the slice width suggests. This finding has practical implications for struc tural MRI and could be used to provide "free" SNR.