A quick and robust method for measurement of signal-to-noise ratio in MRI

A novel method is proposed for the measurement of signal-to-noise ratio (SNR) for the purpose of quality assurance (QA) in MRI. A boxcar filtering technique is applied which allows estimation of signal and noise from a single image. The method has been used to estimate SNR in a large set of images acquired in a consistent manner using various scanner models, coils and phantoms. Performance is evaluated by comparison with the double-image subtraction technique incorporating temporal instability correction. The limits of agreement between the techniques are comparable to typical variability in daily SNR, and significantly superior to the performance of other single-image methods published to date. Single-image methods are preferable as they halve the image acquisition time of the recommended double-image approach. Major inaccuracies are identified in methods of SNR measurement currently used for QA in MRI.