The importance of correcting for signal drift in diffusion MRI

• Published in: Magnetic resonance in medicine 2017-01, Vol.77 (1), p.285-299
• Main Authors: Vos, Sjoerd B., Tax, Chantal M. W., Luijten, Peter R., Ourselin, Sebastien, Leemans, Alexander, Froeling, Martijn
• Format: Article
• Language: English
• Subjects: Anisotropy; artefact correction; Brain - diagnostic imaging; Data analysis; Diffusion effects; Diffusion Magnetic Resonance Imaging - methods; diffusion tensor imaging; Diffusion Tensor Imaging - methods; Drift; Eigenvalues; Eigenvectors; fiber tractography; high angular resolution diffusion imaging; Humans; Image acquisition; Image Processing, Computer-Assisted - methods; Kurtosis; Magnetic resonance imaging; Medicine; Nerve Fibers - physiology; Parameter estimation; Phantoms, Imaging; Resonance; Scanners; Signal Processing, Computer-Assisted; Stability analysis; Tensors
• ISSN: 0740-3194; 1522-2594
• DOI: 10.1002/mrm.26124

Purpose To investigate previously unreported effects of signal drift as a result of temporal scanner instability on diffusion MRI data analysis and to propose a method to correct this signal drift. Methods We investigated the signal magnitude of non‐diffusion‐weighted EPI volumes in a series of diffusion‐weighted imaging experiments to determine whether signal magnitude changes over time. Different scan protocols and scanners from multiple vendors were used to verify this on phantom data, and the effects on diffusion kurtosis tensor estimation in phantom and in vivo data were quantified. Scalar metrics (eigenvalues, fractional anisotropy, mean diffusivity, mean kurtosis) and directional information (first eigenvectors and tractography) were investigated. Results Signal drift, a global signal decrease with subsequently acquired images in the scan, was observed in phantom data on all three scanners, with varying magnitudes up to 5% in a 15‐min scan. The signal drift has a noticeable effect on the estimation of diffusion parameters. All investigated quantitative parameters as well as tractography were affected by this artifactual signal decrease during the scan. Conclusion By interspersing the non‐diffusion‐weighted images throughout the session, the signal decrease can be estimated and compensated for before data analysis; minimizing the detrimental effects on subsequent MRI analyses. Magn Reson Med 77:285–299, 2017. © 2016 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine.