Compressed sensing on DTI via rotating interpolation

Diffusion tensor imaging is a special magnetic resonance imaging method and is widely used to characterize tissue micro architecture and brain connectivity. However, DTI requires long acquisition time due to the repetitive examination with different diffusion gradients and is easily suffered from motion artifacts. These drawbacks greatly limit the clinical application of DTI. In this study, we proposed a rotating interpolated compressed sensing approach. This method rotated sampling masks of each diffusion direction and utilized the k-space data of other diffusion directions in multi-gradient acquisition. The missing k-space data of a highly undersampled image were compensated by the raw data of other diffusion tensor images. Simulations in vivo brain images indicated that the proposed method can further reduce raw data size and enhance the imaging speed without significant sacrifice of image quality and edge information of multi diffusion tensor images over conventional CS methods.