To design a robust, multi-site, multi-vendor protocol to acquire high resolution structural, functional and diffusion MRI data in a time efficient manner.Over the past several years, the Human Connectome Project has pioneered a number of advanced MRI acquisition techniques, allowing for the acquisition of high resolution structural and diffusion MRI scans, as well as fMRI scans with high spatial and temporal resolution. Unfortunately, such advancements have to this point only been available on Siemens scanners. Therefore, we are attempting to extend the benefits of Connectome-style acquisitions to other major scanner manufacturers. This goal is particularly relevant for the recently funded Adolescent Brain and Cognitive Development Study (ABCD) which aims to scan over 10,000 children ages 9-10 and follow them longitudinally for ten years, with each subject getting scanned every 2 years. Creating a standardized acquisition protocol that can be performed on multiple scanner platforms provides maximum utility not only for this ambitious study, but for additional studies that wish to benefit from the enhanced capabilities of these Connectome-like protocols but lack the required Siemens hardware. This abstract will focus on the implementation of what we are calling the Harmonized Human Connectome Protocol for General Electric MRI scanners, specifically the Discovery MR750 scanner, though a similar effort has been completed for Philips scanners. To ensure maximum applicability, this protocol does not require any non-commercially available upgrades. The protocol consists of: 1) a 3D T1-weighted MPRAGE scan, using prospective motion correction (PROMO)¹; 2) a 3D T2-weighted variable flip angle fast spin echo scan, also using PROMO; 3) a high angular resolution diffusion imaging scan, with multiple b-values, and integrated B0 distortion correction (EPIC)²; 4) high spatial and temporal resolution resting state and task fMRI scans, with integrated distortion correction. These pulse sequence parameters are virtually identical to those used in the Lifespan Human Connectome Project, with the exception of the diffusion scan, which has been modified with a larger FOV to minimize phase wrap, and also acquires data at an additional b-value to better allow for the application of a broader set of post-processing methods, including bi-tensor fitting and Restriction Spectrum Imaging. To ensure even greater compatibility across sites and manufacturers for multi-site studies, sites can also participate in a rigorous cross-site calibration procedure that allows for quantitative characterization of the effect of sequence parameters, such as inversion time, echo time, flip angle, and b-value, on imaging signals for each scan modality. This will allow fitting of parameterized physics-based models for each scanner, and enable more advanced cross-scanner (and cross-time) correction procedures, even in the presence of technological change. Additionally, as with the ABCD study, sites can also be provided with the infrastructure to save raw k-space data, which will allow for the application of future improvements in reconstruction technology to previously acquired data.

Protocol development consisted of iterative scan sessions on GE, Siemens and Philips MRI scanners. Using the Lifespan Human Connectome protocol as a starting point, changes were introduced to improve image quality, both in raw images as well as error-corrected post-processed images. The Harmonized Human Connectome Protocol for GE (see Table 1 for detailed scan parameters) differs mostly in the diffusion scan, with the expansion of the FOV and the inclusion of an additional b-value.

Quantitative models (Figure 2) were generated based on multiple diffusion acquisitions on a GE Discovery MR750, with varying TE and b-value. Diffusion data were fit to a bi-tensor model, with separate pools for free and restricted diffusion, each with its own R2.

Figure 1 shows multiband diffusion acquisitions for Siemens (Prisma 3T) and GE (Discovery MR750) scanners in the same subject. These scans show good qualitative similarity, despite being processed without any site normalization or calibration techniques. However, in addition to this qualitatively similar starting point, Figure 2 illustrates our quantitative cross-site normalization procedures as applied to diffusion MRI, showing an excellent fit of observed data as a function of b-value and TE with a multi-compartmental mixture model, and their predicted effects on the derived ADC and FA parameters. The resulting quantitative signal models and calibration parameters can then be incorporated into the image post-processing methods, to minimize cross-site effects.A robust, high-resolution multiband protocol, similar to the Lifespan Human Connectome protocol, which was previously only available on Siemens scanners, is possible on a GE MR750 with commercially available hardware packages. In addition, the continued development of quantitative calibration procedures should further reduce site/manufacturer differences in derived diffusion measures for large, multi-site, multi-vendor studies such as the Adolescent Brain and Cognitive Development study.