Noise and texture characteristics of selective averaging

Selective averaging, where different lines in k-space are accumulated with a different number of acquisitions has previously been shown to introduce texture and mottling into magnetic resonance (MR) images. Here a theoretical framework for understanding and quantifying the noise characteristics of selective averaging is presented. In this analysis the noise power spectrum is used to predict the value of the noise and its texture by considering the granularity function. Acquired and simulated MR images are used to verify the accuracy of these theoretical predictions. Selective averaging increases signal/noise but with significantly reduced time efficiency compared to nonselective acquisitions. The colored quality of the noise with selective averaging can introduce unwanted artifacts. In this case the relative improvement in signal detection efficiency as imaging time is increased will be object size dependent