Reliable cortical thickness estimation with reduced susceptibility-induced signal loss and inherent blood suppression in single-slab 3D turbo spin echo imaging

Abstract

The thickness of the human cerebral cortex, which provides valuable information in the studies of normal and abnormal neuroanatomy, is commonly estimated using high-resolution, volumetric magnetization-prepared rapid gradient echo (MP-RAGE) magnetic resonance imaging. MP-RAGE is widely accepted in the neuroanatomical studies due to its strong T1-weighted contrast and high signal-to-noise ratio. However, since MP-RAGE is based on free-induction-decay (FID) signals and thus potentially leads to susceptibility-induced signal loss, it is often challenging to accurately estimate cortical thickness. Additionally, blood vessels, which appear with high signal intensity due to flow-related enhancements, may simulate white matter structures, potentially impairing brain segmentation in image processing. In this work, for accurate and reliable estimation of cortical thickness of the human brain we develop and optimize T1-weighted, single-slab three-dimensional (3D) turbo/fast spin echo (SE) imaging, wherein 1) instead of FID signals radio-frequency pulse refocused echo signals, which are resistant to susceptibility-induced signal loss, are acquired, 2) variable low flip angles (VFA) in the refocusing pulse train are utilized, increasing the incoherence of SE and stimulated echo signals particularly for blood vessels and thus leading to inherent blood suppression, and 3) a composite flip-down pulse is employed at the end of the refocusing pulse train, generating partial inversion recovery and thereby enhancing T1-weighted contrast. Numerical simulations of the Bloch equation are performed in both the MP-RAGE and proposed method for comparison, investigating the effect of susceptibility variations on signal evolutions along the echo train for stationary tissues and estimating the effect of the VFA and increased spoiling scheme on blood suppression. In vivo whole-brain data are acquired in fourteen volunteers. Image processing is then performed using the Freesurfer, calculating mean and standard deviations of cortical thickness for the entire cortical surfaces and investigating susceptibility- or high-intensity-blood-vessel-induced segmentation errors. Statistical analysis demonstrates that particularly in the inferior prefrontal and medial temporal regions affected by rapid susceptibility variations the MP-RAGE imaging, if compared with the proposed method, significantly under-estimates cortical thickness. Thus, in the studies of brain morphometry the proposed method is potentially a promising alternative to conventional MP-RAGE.