Developing next-generation imaging technologies for in-vivo mesoscale diffusion MRI and microstructure imaging - PROJECT SUMMARY / ABSTRACT Diffusion MRI (dMRI) is an important non-invasive imaging tool in both clinical practice and neuroscientific research. However, current in-vivo dMRI i) lacks sensitivity to detect small yet crucial structures, which not only constitute a significant portion of the whole-brain connectivity, but also hold critical clinical importance. For example, the small deep brain fibers are critical targets in deep brain stimulation that are used to treat many diseases; and ii) has limited sensitivity to measure tissue microstructure, which provides valuable complementary information to high-resolution dMRI and offers new biomarkers for early diagnosis and treatment assessment. To address this, high spatial resolution, high b-value, and/or rich information content are imperative. However, significant technical challenges persist in achieving these within practical scan time, including low SNR, high vulnerability to motion, and detrimental geometric distortion, blurring, and slab-boundary striping artifacts. In this proposal, we will develop next-generation in-vivo diffusion MRI technologies that overcome critical challenges that even state-of-the-art methods have fallen short in addressing so far, and enable motion-robust and high-quality mesoscale-resolution dMRI and high b-value microstructure imaging in a short scan time. Specifically, we will develop novel acquisition technologies that can provide significant SNR gain (√25 folds, 25× scan time reduction), high robustness to various sources of motion, and unprecedented image quality completely free from detrimental distortion, T2/T2* blurring, and slab-boundary artifacts. They will also address the long- standing challenge—vulnerability to motion/eddy-current induced dynamic distortions—thereby achieving superior image sharpness and enabling motion-robust joint acceleration. Our preliminary results demonstrate that it enables, for the first time ever, mesoscale in-vivo dMRI on both 3T and 7T clinical scanners (485 μm-iso). We will further develop and refine these acquisition technologies, and integrate them with novel synergistic motion-aware joint reconstruction and PCA-based or self-supervised denoising techniques, to enable a remarkable overall 100-fold improvement in scan speed. This will enable in-vivo mesoscale dMRI, and microstructural dMRI with rich information content (high b, multiple diffusion times, and simultaneous multi-TEs), to be acquired within clinically-feasible scan times. In addition to disseminating our sequences, reconstruction, and data developed on broadly available 3T and 7T clinical scanners, we will create and share benchmark in- vivo dMRI datasets acquired on the ultra-high-performance Connectome 2.0 scanner, matching the quality and resolution of the state-of-the-art ex-vivo human brain data. We believe the proposed technologies will transform the current diffusion imaging and its neuroscientific and clinical applications.
