Reduced Encoding Diffusion Spectrum Imaging Implemented With a Bi-Gaussian Model

Abstract

Diffusion spectrum imaging (DSI) can map complex fiber microstructures in tissues by characterizing their 3-D water diffusion spectra. However, a long acquisition time is required for adequate q-space sampling to completely reconstruct the 3-D diffusion probability density function. Furthermore, to achieve a high q-value encoding for sufficient spatial resolution, the diffusion gradient duration and the diffusion time are usually lengthened on a clinical scanner, resulting in a long echo time and low signal-to-noise ratio of diffusion-weighted images. To bypass long acquisition times and strict gradient requirements, the reduced-encoding DSI (RE-DSI) with a bi-Gaussian diffusion model is presented in this study. The bi-Gaussian extrapolation kernel, based on the assumption of the bi-Gaussian diffusion signal curve across biological tissue, is applied to the reduced q-space sampling data in order to fulfill the high q-value requirement. The crossing phantom model and the manganese-enhanced rat model served as standards for accuracy assessment in RE-DSI. The errors of RE-DSI in estimating fiber orientations were close to the noise limit. Meanwhile, evidence from a human study demonstrated that RE-DSI significantly decreased the acquisition time required to resolve complex fiber orientations. The presented method facilitates the application of DSI analysis on a clinical magnetic resonance imaging system.