Precision Brain Morphometry: Feasibility and Opportunities of Extreme Rapid Scans [article]

Jared A. Nielsen, Ross W. Mair, Justin T. Baker, Randy L. Buckner
2019 bioRxiv   pre-print
The traditional approach to achieve optimal structural brain morphometry is to increase spatial resolution and maintain a high signal-to-noise ratio within a single scanning session using long acquisitions. Here we explore alternative approaches that use multiple rapid structural scans and averaging to achieve stable morphometric estimates. Effects of sequence, resolution, acceleration, and number of scans within- and between-sessions were varied across six studies that included 1,025 scans.
more » ... gle ~1-min magnetization-prepared rapid gradient echo (MP-RAGE) scans that use wave-controlled aliasing in parallel imaging (Wave-CAIPI) encoding yield morphometric estimates nearly as accurate as those obtained with traditional long scans. Averaging estimates from multiple rapid scans reduces error below traditional approaches using less total acquisition time. Test-retest reliability within- and between-sessions for ~2 min multi-echo MP-RAGE scans revealed that a source of error is the idiosyncratic positioning of the subject's head between sessions. Cluster scanning over multiple tightly-spaced sessions mitigates this effect and achieves cortical thickness estimates with less than 10μm global (30μm regional) error. Test-retest error for the hippocampus falls below 0.6%. Additional sources of error include coil heating when scans are acquired rapidly within the same session and change in console software between sessions. Rapid acquisitions allow novel experimental designs that can use cluster scanning to achieve precise longitudinal assessment within the individual, build in robustness by affording redundant acquisitions, and reduce the structural scanning burden to a minute when anatomical registration is the goal. The field should consider replacing long structural scans with fast alternatives.
doi:10.1101/530436 fatcat:ve5nbxxwvrcjhjeeetq2lotsbq