Time‐efficient and flexible design of optimized multishell HARDI diffusion

    loading  Checking for direct PDF access through Ovid

Excerpt

Diffusion magnetic resonance imaging (dMRI) is able to provide insight into the complex neural fibre tract architecture of the human brain in vivo 1. It has clinical applications in the study, diagnosis and treatment of neurological disorders 2, as well as many scientifically focused applications, with an important emerging focus on building comprehensive models of human brain connectivity 3. The acquired dMRI data, using specific gradients with varying magnitude and direction that introduce sensitivity to water diffusive processes of molecules in the brain, provide the input to the chosen analysis pipeline and thus determine and limit its capacity to probe structural complexity of the human brain. The potential of dMRI in this context continues to grow with advanced techniques such as biophysical modelling 5, which allow more precise mapping of micro‐structure. However, these advanced analyses require ever more data, with current trends requiring not only the sampling of a large number of diffusion sensitization directions, such as in high angular resolution diffusion imaging (HARDI) 9, but also the collection of multiple b‐value shells 6. Furthermore, strategies such as diffusion spectrum imaging 10 focus on the whole diffusion sensitization space, constructed by diffusion directions and sensitization levels beyond the single spherical shell concept. It is clear that the growing range of diverse analysis models motivates a drive toward both increased time efficiency and flexibility regarding the choice and ordering of diffusion sensitization in the dMRI acquisition.
An ideal dMRI acquisition would provide whole brain coverage volumes with a high spatial resolution, combined with a completely flexible sampling of the diffusion sensitization space while producing data with high signal‐to‐noise ratio (SNR) at a maximum rate of data acquisition to allow as many independent samples as possible in a feasible examination time. However, both physiological and hardware constraints impose limits. Diffusion scans with the desired high sensitization (b‐value) push the boundaries of the capacities of modern gradient coil systems, with maximum gradient amplitude, slew rate and duty cycle (fraction of time for which demand can be placed on gradient system) all imposing limits on the acquisition. In addition to hardware constraints, there are also challenges associated with in vivo studies, such as limited scan time, subject motion, scan interruptions, and even early scan aborts 11. To maximize utility, the acquisition needs to be carefully designed to be as time‐efficient and self‐consistent as possible and ideally to be robust to non‐ideal examination subjects and conditions. These requirements hold for any clinical cohort, but are particularly important for studies of neonates. The analysis of 52 dMRI scans from non‐sedated neonatal term subjects (gestational age 37.28–46.57 weeks, median 41.07 weeks) imaged in our institution in 2013 reveals that the acquisition was stopped before completion 17 times (33%) and required parts to be repeated in another 14 instances (27%) due to awakening or motion. Measures to accommodate this are thus beneficial for any study. The stimulus for this work is the developing Human Connectome Project (dHCP),1 aspiring to collect data from over 1000 newborns to gain insight into the development of the human brain. For the dHCP, the dMRI acquisition needs to be completed in approximately 20 min. The data will be shared with the scientific community and should thus be suitable for a wide range of processing algorithms and neuro‐developmental studies.
With an acquisition time (TA) per slice shorter than 100 ms and tolerance to background phase variation caused by bulk motion 13, single‐shot spin‐echo echo‐planar imaging is still the most commonly used imaging method for dMRI. However, echo‐planar imaging (EPI) is sensitive to both T2 and JOURNAL/mrim/04.02/01445475-201803000-00007/math_7MM1/v/2018-01-24T161827Z/r/image-png decay.

Related Topics

    loading  Loading Related Articles