Introduction

The gray matter (GM) microarchitecture is generally organized into vertical (cortical columns) and horizontal units (cortical layers). In humans, the cortical columns are typically 0.6-0.8 mm in length, with similar dimensions observed for cortical layers (i.e. six layers across 3-4 mm cortical thickness). These columns and layers are specifically organized to play important roles in brain function and development, and abnormalities in these microstructures can directly affect brain functions, well before the symptoms occur. As such, the characterization of cortical columns that traverse the layers of GM is essential in providing early imaging biomarkers for neurological disorders. In an effort to delineate these fine-grain GM microarchitectures, diffusion tensor imaging (DTI) at extremely high spatial resolutions (<0.5 mm) have been obtained ex-vivo¹. To achieve such high spatial resolutions in-vivo, however, challenges remain due to technical limitations (i.e. single-shot EPI used for human DTI in vivo) and physiological confounds (e.g. brain pulsation). This study aims to develop robust and effective solutions to reach the 0.6-0.8 mm isotropic spatial resolution necessary to achieve a detailed delineation of human brain GM in-vivo, while remaining within the limits of conventional imaging times on a 3T MRI scanner.

Methods

In-vivo DTI data was acquired in a GE (Waukesha, WI) Premier Performance 3 T scanner equipped with a high-power 60cm gradient coil (peak strength 115 mT/m) specifically designed for efficient diffusion MRI. A multiplexed sensitivity-encoded (MUSE)² DTI acquisition was employed with 4 interleaved excitations. Written informed consent was obtained from human subjects in accordance with our institutional IRB. It has been shown that DTI data with a 0.8 mm isotropic spatial resolution has sufficient SNR to achieve reliable fiber tractography in white matter3using a b-factor of 800 s/mm². To determine the appropriate b-factor for GM imaging (which is hypothesized to be more suitable for single-tensor fiber tracking models given its well-organized columnar structure), DTI datasets with 15 diffusion directions were first acquired with a 0.8 mm isotropic spatial resolution at a b-factor of 800 s/mm², and then at lower b-factors of 600 and 400 s/mm² to improve the signal-to-noise ratio (SNR). After reconstructing each multi-shot data set with MUSE, MRtrix3⁴ was used to remove noise and eddy current effects, apply a bias field correction, generate diffusion tensors, and finally derive deterministic streamline fiber tracts. By limiting the fiber pathway length to <3 mm, short cortical fibers were delineated specifically in the GM. Using a single-tensor deterministic tracking model⁵, it was confirmed that that a b-factor of 800 s/mm²was also optimal for tracking GM cortical columns, despite the fact that DTI images at b-factors of 600 and 400 s/mm² had higher SNR. Using this optimized b-factor of 800 s/mm², DTI at an even higher spatial resolution of 0.6 mm was acquired in the same human subjects using the same 15 diffusion directions. To achieve sufficient SNR for the reduced voxel volume, four such datasets were acquired and averaged. Subsequently, the same post-processing and fiber tracking procedures were performed.

Results and Discussion

Fig. 1 presents all fibers under 3 mm in length in an axial slice of the brain. It is apparent that the vast majority of these short fibers reside within the GM and upon close inspection in the enhanced regions, the fibers span across the GM from the white matter boundary to the outer most cortical layer. As these fibers are derived from voxels 0.6 mm in dimension, they can be used to characterize the structural makeup of the cortical layers, ultimately highlighting regions with abnormalities that associate with neurological disorders.

Conclusion

Although cortical columns have been observed at ultra-high spatial resolutions in ex-vivo data, the attainable spatial resolution for delineating these structures through DTI in-vivo has long been limited by multiple technical constraints. In this study, it has been shown that ultra-high spatial resolution DTI can be realized in-vivo to effectively resolve cortical columns in GM. An ability to accurately characterize these columns can be used to derive imaging biomarkers to detect the early onset of neurological disorders.

Acknowledgements

This work was funded in part by NIH grants R01 NS 075017, R24-106048, and S10-OD-021480.