Pushing the limits of in vivo diffusion MRI for the Human Connectome Project

Abstract

Perhaps more than any other "-omics" endeavor, the accuracy and level of detail obtained from mapping the major connection pathways in the living human brain with diffusion MRI depend on the capabilities of the imaging technology used. The current tools are remarkable; allowing the formation of an "image" of the water diffusion probability distribution in regions of complex crossing fibers at each of half a million voxels in the brain. Nonetheless our ability to map the connection pathways is limited by the image sensitivity and resolution, and also the contrast and resolution in encoding of the diffusion probability distribution. The goal of our Human Connectome Project (HCP) is to address these limiting factors by re-engineering the scanner from the ground up to optimize the high b-value, high angular resolution diffusion imaging needed for sensitive and accurate mapping of the brain's structural connections. Our efforts were directed based on the relative contributions of each scanner component. The gradient subsection was a major focus since gradient amplitude is central to determining the diffusion contrast, the amount of T2 signal loss, and the blurring of the water PDF over the course of the diffusion time. By implementing a novel 4-port drive geometry and optimizing size and linearity for the brain, we demonstrate a whole-body sized scanner with G(max) = 300 mT/m on each axis capable of the sustained duty cycle needed for diffusion imaging. The system is capable of slewing the gradient at a rate of 200 T/m/s as needed for the EPI image encoding. In order to enhance the efficiency of the diffusion sequence we implemented a FOV shifting approach to Simultaneous MultiSlice (SMS) EPI capable of unaliasing 3 slices excited simultaneously with a modest g-factor penalty allowing us to diffusion encode whole brain volumes with low TR and TE. Finally we combine the multi-slice approach with a compressive sampling reconstruction to sufficiently undersample q-space to achieve a DSI scan in less than 5 min. To augment this accelerated imaging approach we developed a 64-channel, tight-fitting brain array coil and show its performance benefit compared to a commercial 32-channel coil at all locations in the brain for these accelerated acquisitions. The technical challenges of developing the over-all system are discussed as well as results from SNR comparisons, ODF metrics and fiber tracking comparisons. The ultra-high gradients yielded substantial and immediate gains in the sensitivity through reduction of TE and improved signal detection and increased efficiency of the DSI or HARDI acquisition, accuracy and resolution of diffusion tractography, as defined by identification of known structure and fiber crossing.

Keywords: DSI; Diffusion imaging; Gradient hardware; HARDI; MRI; Structural connectivity.

Published by Elsevier Inc.

Figures

Fig. 1

Gradient drive configuration for the 300mT/m Connectome gradients. Gy axis shown in cartoon form. The high inductance needed to achieve the needed gradient strength is split into 4 sections, each consisting of a “finger-print” primary coil and its associated shield. Thus the inductance of each section is comparable to a conventional whole body gradient (~1mH) and can be switched rapidly by the 2000V gradient amplifiers. Driving each axis with 4 parallel drives requires both a high degree of synchronization and a matrix style configuration of the gradient regulators needed to control overshoots.

Fig. 2

Photo of the completed gradient set in the bore. Note the wide (14cm) annulus of the gradients due to the decreased inner diameter (610mm) and full-sized outer diameter (890mm).

Fig. 3

Peripheral nerve stimulation thresholds for the Gy axis of Connectome gradient compared to a conventional 45mT/m whole body gradient. The reduced linearity of the Connectome gradient (5% deviation from linear on a 20cm FOV) reduces the maximum B field excursion created by the gradient and thus lowers dB/dt and nerve stimulation. This allows an improved EPI readout. For example, the Connectome gradient could achieve an 18% reduction in the EPI echospacing (and thus image distortion) without nerve stimulation compared to the conventional gradient.

Fig. 4

Peripheral nerve stimulation as well as cardiac thresholds for the Connectome gradient. The hardware limits can easily surpass either of these limits and therefore the system monitors the gradient waveforms and stops the scan if either limit is exceeded.

Fig.5

Left) Minimum TE obtained for a standard Skeskjal Tanner spin echo diffusion sequence as a function of b value as a function of maximum gradient strength. Results are for a 2mm isotropic EPI readout (200 FOV). Right) Measured SNR of the brightest sections of in vivo human white matter (where fiber orientation is orthogonal to the applied diffusion gradient) as a function of maximum gradient strength for the same acquisition at b = 10,000s/mm2 and 20,000 s/mm2. SNR is normalized to the SNR obtained with Gmax = 40mT/m. Data for 5b acquired with the 64ch brain array.

Fig. 6

Diffusion weighted images (single direction) acquired at Gmax = 40mT/m, 100mT/m and 300mT/m with b=10,000 s/mm2 and 1.5mm isotropic resolution. Data acquired with the 64ch brain array.

Fig. 7

Design, implementation and testing of the 64ch brain coil for the Connectome scanner. Layout of the circular receive elements is shown on the two halfs of the former, as well as the finished coil with and without covers and its relative SNR gain compared to a sized matched 32ch array for accelerated brain imaging as measured in a peripheral, intermediate and central brain ROI.

Fig. 8

Blipped-CAIPI acquisition scheme for MB2 and FOV/2 PE shift. The additional Gz encoding gradients applied simultaneously with the standard Gy phase gradient in the EPI readout are shown on the left. Each Gz gradient blip causes a π phase change in the signal of the top imaging slice. This results in a phase modulation that is equivalent to a linear phase causing a desired FOV/2 shift as shown on the right. The reversal of every other Gz blips minimizes the accrual of intravoxel dephasing along the slice direction and the associate voxel tilting artifact.

Fig. 9

Blipped-CAIPI retained SNR results for MB factor of 3 (Top) providing ~100% of the unaccelerated SNR. Middle) 3 fold accelerated q-ball. Bottom) 3 fold accelerated DSI () acquisitions. All results used the 64ch brain array.

Fig. 10

Leaked and blocked signal of slice 4 in a MB-5 blipped-CAIPI acquisition from i) standard Slice-GRAPPA (Std-SG) and ii) Split Slice-GRAPPA (Sp-SG) reconstructions. The Sp-SG reconstruction results in a significant reduction in leaked and blocked signal artifacts.

Fig. 11

Top) Tractography results from a 4-minute DSI scan acquired using MB-3 and 4 fold Q-space compressed sensing (CS). Bottom) Tractography results and average FA values over 18 major white matter pathways from i) fully sampled 515 DSI with MB-3 (16 minutes) and ii) 4 fold Q-space compressed sensing (CS) DSI with MB-3 (4 minutes). Good agreements of results from these two datasets can be observed. The 64 channel brain array was used.

Fig. 12

Sagittal single shot SE-EPI dataset showing level of susceptibility induced geometric distortion present in the Connectome scanner data. This is the b=0s/mm2 data from a b=10,000s/mm2 dataset acquired with 1.5mm isotropic resolution and R=3 in-plane GRAPPA acceleration. In addition to the distortion mitigation of the R=3 GRAPPA, the data is also mitigated by the faster readout of the Connectome gradient. Slices from half the brain are shown to improve visibility.

Fig. 13

Residual bootstrap analysis of 95% confidence interval of the angular uncertainty of the second fiber direction in b = 10,000s/m2, 160 directions, 1.5mm isotropic resolution diffusion data. The 64ch brain array was used.

Fig. 14

Comparison of diffusion tractography at Gmax = 40, 100 and 300 mT/m from a Q-Ball type acquisition (1.5 mm isotropic, 160 directions, b-value 10,000 s/mm2). TEs were 100 ms, 66 ms and 54ms for Gmax = 40 mT/m, 100 mT/m and 300 mT/m, respectively. Higher SNR enables better depiction of U-fibers with fewer false positives. The 64ch brain array was used.