Diffusion MRI (dMRI) is used to quantify brain tissue microenvironments and obtain accurate estimates of axonal orientations for tractography by utilizing the characteristics of the molecular diffusion of water. Precise acquisition of dMRI images requires probing water molecules in multiple diffusion scales and directions, which consequently increases acquisition time. Increased acquisition times are especially vulnerable to motion artifacts and limits the use of dMRI in patients who cannot remain relatively motionless, such as pediatric and Parkinson’s disease patients. Conventional methods of accelerated acquisition rely on reduction of the repetition time, which causes loss of signal fidelity and is prone to artifacts. This demonstrates the need for improved methods of accelerated image acquisition.
Researchers in the Department of Radiology and the Biomedical Research Imaging Center have developed a novel dMRI sampling and reconstruction technique termed Slice-Interleaved Diffusion Encoding (SIDE), which accelerates dMRI image acquisition by encoding slices in an image volume with multiple diffusion gradients. SIDE accelerates acquisition by 50-fold when combined with simultaneous multislice (SMS) techniques of factor 5. Using the SIDE technique, each SMS slice group in an image volume is excited with a different diffusion wavevector, generating a multitude of slice-under-sampled diffusion weighted (DW) images. A full DW image is then generated from the under-sampled slice groups using a graph convolutional neural network. The proposed graph convolutional neural network based reconstruction method provides more structural details in generated images than current techniques. Reducing dMRI acquisition time using SIDE allows as many as 200 DW images, at high 1.5 mm isotropic resolution and with whole-brain coverage, to be acquired in one minute. This massive improvement in acquisition speed will considerably improve the clarity of images and allow dMRI to be utilized for patients that cannot remain relatively motionless for extended periods of time, such as pediatric patients.
- Marked reduction in acquisition time of dMRI data (50-fold decrease when combined with SMS factor 5).
- SIDE acquires more incoherent data without reducing the repetition time, thereby preventing reduction in the signal-to-noise-ratio and preventing unwanted artifacts.
- Provides better structural details than current reconstruction techniques.
- Reduced computational complexity and memory requirements using undersampled DW images fed into graph convolutional neural network.
SIDE can be applied to improve the acquisition speed and image quality in dMRI.
- Multifold Acceleration of Diffusion MRI via Slice-Interleaved Diffusion Encoding (SIDE) Yoonmi Hong, Wei-Tang Chang, Geng Chen, Ye Wu, Weili Lin, Dinggang Shen, Pew-Thian Yap. arXiv:2002.10908 [physics.med-ph]