Improving the sensitivity and specificity of diffusion MRI

PROJECT SUMMARY/ABSTRACT The ability to detect fiber pathways and assess tissue microstructure in vivo has opened a new window on the brain, with important applications ranging from brain connectomics to stroke detection and surgical planning. Diffusion Magnetic Resonance Imaging (DMRI) has the unique ability to reveal anatomical connectivity and tissue microstructure using noninvasive imaging methods. The development of open-source software packages for DMRI analysis has fueled the rapid growth of applications in both systems neuroscience and clinical neuroimaging. However, these applications have raced far ahead of the validation effort required to establish the reliability of the methods and quantify the impact of basic assumptions made by competing algorithms. For example, simplifying assumptions in commonly used analysis packages preclude detailed tissue characterization, ignoring untapped information in the diffusion weighted MRI signal. On the other hand, many microstructural analysis methods ignore fiber dispersion, which is prevalent throughout the white matter. The goal of this project is to quantify and improve the reliability of sub-voxel measurements of fiber properties by simultaneously improving angular resolution and sensitivity to fiber-specific diffusion properties. The project has 3 specific aims. The first aim is to develop and validate improved methods for sub-voxel fiber identification and tissue characterization. Results of our previous studies show that current methods have limited ability to resolve complex fiber distributions when crossing angles are less than ~40-60° (depending on data acquisition parameters). This limit biases fiber tractography and precludes the possibility of accurate fiber-specific microstructure measurements. We will compare the ability of our new method and current leading algorithms to segment and characterize sub-voxel fibers, using confocal microscopy data from the squirrel monkey as ground truth. The second aim is to determine the reproducibility of fiber-specific diffusion properties, both through space (uniformity along a pathway) and time (test-retest reproducibility) and the dependence of these properties on tissue microstructure, as measured by optical microscopy. These data will be used in critical tests of key assumptions made by analysis algorithms in current use. The third aim is to quantify the intersubject variability of fiber-specific diffusion properties, both in squirrel monkeys and human subjects. We will expand our squirrel monkey brain atlas to include the new fiber-specific diffusion properties and test the hypothesis that these measures are highly reproducible across healthy individuals, as suggested by our preliminary data. Creation of a similar human atlas will provide a framework for testing the hypothesis that intersubject variability of fiber-specific diffusion properties is lower than that of current DMRI measures, which do not fully account for fiber dispersion or crossing. In combination, these aims have the potential to improve the sensitivity of DMRI to white matter injury, while providing a clear link between the imaging biomarkers and biophysical properties of the tissue, thereby improving the specificity of DMRI in detecting changes due to injury, recovery, and aging.