Simultaneous Bayesian correction of slab boundary artifacts and bias field for high resolution ex vivo MRI

Abstract

Multi-slab MRI overcomes some of the hardware limitations of today's clinical scanners (e.g., memory size), enabling the acquisition of ultra-high resolution ex vivo MRI of the whole human brain with high SNR efficiency. However, multi-slab MRI suffers from slab boundary artifacts (SBA) that can greatly bias subsequent analyses. Since SBA heavily interplays with the bias field (BF) present in MRI, we propose a Bayesian method that corrects for SBA and BF simultaneously. The method, which combines a probabilistic brain atlas with an Expectation Maximization inference algorithm, is shown to outperform state-of-the-art SBA and BF correction techniques - even when used in combination.