Signal-to-noise ratio of diffusion weighted magnetic resonance imaging: Estimation methods and in vivo application to spinal cord

Diffusion tensor imaging (DTI) and tractographic reconstruction may be applied for in vivo clinical spinal cord studies. However, this structure represents a challenge to current acquisition and reconstruction strategies, due to its small size, motion artifacts, partial volume effects and low signal-to-noise-ratio (SNR). Aims of this work were to select the best approach for the estimate of SNR and to use it for spinal cord diffusion weighted (DW) sequence optimization.Seven methods for the estimate of SNR were compared on uniform phantom DW images, and the best performing approach (single ROI for signal and noise, difference of images—SNRdiff) was applied for the following in vivo sequence evaluations.Fifteen sequences with different parameters (voxel size, repetition (TR) and echo (TE) times) were compared according to SNR, resolution, fractional anisotropy (FA) and tractography performances on three healthy volunteers. In vivo optimization of DW sequences resulted in: axial sequence, with voxel size = 1.5 mm × 1.5 mm × 3.5 mm, TR = 3200 ms and TE = 89 ms, sagittal sequence with voxel size = 2.2 mm × 2.2 mm × 2 mm, TR = 3000 ms and TE = 84 ms.An objective method tested on phantom and a practical index for in vivo spinal cord DTI SNR estimation allowed to obtain axial and sagittal optimized sequences, providing excellent tractographic results, with acceptable acquisition times for in vivo clinical applications.

► Standard deviation of the difference image from two acquisitions best estimates noise. ► Some SNR estimation methods are inapplicable in in vivo spinal cord studies. ► Many factors have to be considered in diffusion weighted sequence optimization. ► A quality index is proposed as useful tool in selecting and comparing DW MR sequences.