Rapid 1 mm isotropic diffusion tensor imaging with denoising and improved parameter estimation for detecting focal hippocampal lesions in temporal lobe epilepsy

Abstract

While high resolution diffusion tensor imaging (DTI) at 1 mm isotropic can detect focal lesions of the hippocampus in temporal lobe epilepsy (TLE), faster acquisition times would facilitate potential clinical implementation. The purpose here is to assess different published denoising algorithms to overcome the low signal-to-noise ratio and accelerate 1 mm isotropic DTI of the human hippocampus at 3 T while maintaining diffusivity metric accuracy and image quality for focal lesion detection in TLE. The previously published 5.5 min protocol of 110 diffusion images per slice (10 directions × 10 averages and 10 b = 0 s/mm²) was assessed for subsets of 1–10 averages (same 10 directions) that were denoised using four algorithms that have been applied to other diffusion MRI datasets. In healthy controls, the variance-stabilizing transformation and optimal singular-value manipulation (VST) and Non-Local Spatial and Angular Matching (NLSAM) denoising greatly improved image quality while minimizing voxels with spurious extremes of fractional anisotropy (FA) or mean diffusivity (MD) down to 4 averages (i.e. 40 diffusion images and 4 b = 0 s/mm²) in healthy controls. The identification of focal lesions indicated by elevated MD and alterations of internal micro-architecture with only 4 averages were comparable to the full data set of 10 averages. Therefore, denoising of 1 mm isotropic DTI of the hippocampus enables a clinically feasible scan time of 2.2 min at 3 T that can be used for the detection of focal hippocampal lesions in TLE, as well as other neurological disorders such as multiple sclerosis, dementia and Alzheimer's disease.