Artefact Correction in DTI (ACID) (ACID) Wellcome Trust Centre for - PowerPoint PPT Presentation

Artefact Correction in DTI (ACID) (ACID) Wellcome Trust Centre for Neuroimaging, UCL Institute of Neurology, University College London Siawoosh Mohammadi

Motivation Potential problems in DTI High-end DTI: tractography y z x Lazar, NMR Biomed., 2010 Mohammadi et al., MRM, accepted

Overview • Diffusion Tensor Imaging (DTI) in brief • Example application in DTI • Three artefacts in DTI • Three artefacts in DTI – Eddy Current (EC) distortions – Local Perturbation Fields (LPFs) – Signal-dropout due to mechanical vibration • Take home message

Diffusion Tensor Imaging (DTI) in brief DT represented Diffusion tensor n DW images by ellipsoid + m b=0 images

Overview • Diffusion Tensor Imaging (DTI) in brief • Example application in DTI • Three artefacts in DTI • Three artefacts in DTI – Eddy Current (EC) distortions – Local Perturbation Fields (LPFs) – Signal-dropout due to mechanical vibration • Take home message

Patients (TLE) and Control Keller et al., Journal of Neuroimaging, accepted

7T – high resolution DTI Heidemann et al., MRM, 2010

Grey matter DTI Variability in grey matter diffusion Amygdala parcellation Nagy et al., ISMRM, 2011 Bach et al., J Neurosci., 2011 Cortical radial and tangential diffusivity MacNab et al., ISMRM, 2011

High angular resolution diffusion imaging (HARDI) ODF - Orientation Distribution Function Aganj et al., MRM, 2010

Overview • Diffusion Tensor Imaging (DTI) in brief • Example application in DTI • Three artefacts in DTI – Eddy Current (EC) distortions – Local Perturbation Fields (LPFs) – Signal-dropout due to mechanical vibration • Take home message

EC distortion artefact Stejskal & Tanner, JCP, 1965 Reese et al., MRM, 2003

EC and imaging gradients z 0     y 0     G EC   z y y   G EC x   x 0     0   y 0   x   G  EC  y   0   Skare S., thesis, 2002

Whole-brain eddy current distortions original image y z y z x y x y distorted image in-plane shearing through-plane shearing scaling translation   G 0 0 EC     eddy current   x     B 0 EC   0 G  EC    field y 0       0 0 G EC       z components Mohammadi et al., MRM, 2010

Eddy currents: bright edges / blurring Without eddy current and With eddy current and motion correction motion correction

Relevance • Less blurring leads to higher sensitivity in FA group comparison Keller et al., JON, accepted • Better tensor estimates • Better tensor estimates towards the cortex improves Nagy et al., ISMRM, GM DTI specificity 2011 • Better image quality in high resolution DTI and HARDI, where ST pulse is necessary Heidemann et al., Aganj et al., MRM, MRM, 2010 2010

Overview • Diffusion Tensor Imaging (DTI) in brief • Example application in DTI • Three artefacts in DTI – Eddy Current (EC) distortions – Local Perturbation Fields (LPFs) – Signal-dropout due to mechanical vibration • Take home message

Problem: effective gradient, e.g., due to ECs diffusion weighting period readout period expected gradients effective gradients EC distortion Error in B matrix FA original FA inhomogeneity

SM2 How to measure the LPFs? Mohammadi et al., Neuroimage, under review

Folie 18 SM2 cite zoltan Siawoosh Mohammadi; 08.11.2011

Measuring LPFs on different MR systems (b) DTI2 (c) DTI3 (a) DTI1 ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε ε 11 22 11 11 22 22 ε ε ε ε + + + + ε 12 ε ε ε ε ε ε ε + + ε ε ε ε ε ε ε ε 12 + + + + + + ε ε ε ε 12 33 33 33 0.1 60 0.1 0.02 60 60 50 50 50 40 40 40 0 0 30 0 30 + + ε ε ε ε 13 + + + + + + 30 ε 23 ε ε ε ε ε ε 13 ε + + + + + + ε ε ε 23 ε + + + + ε ε ε ε 13 + + + + + + ε ε ε ε 23 20 20 20 10 10 10 -0.1 -0.1 -0.02  + +  ε ε ε   11 12 13 B * B B = + δ δ B = 2 Σ + B Σ + = ε + ε ε +   with and 12 22 23   ε + ε + ε   13 23 33 Mohammadi et al., Neuroimage, under review

LPF correction: repositioning experiment 0.1 0.1 z DTI3,2 = 53 ± 3 0.05 z DTI3,1 = 41 ± 3 0 0 −0.05 -0.1 −0.1 ( B ) δ ∆ ∆ cor2 tr cor2 MD MD DTI3,1 DTI3,2 Measured MD Corrected MD number of voxel MD meas number of voxel MD cor2 DTI3,1 DTI3,1 MD meas MD cor2 DTI3,2 DTI3,2 5000 5000 0.5 1 1.5 0.5 1 1.5 mm 2 mm - 3 2 MD [10 ] - 3 s MD [10 ] s Mohammadi et al., Neuroimage, under review

Relevance • Improved sensitivity of group comparison of MD Keller et al., due to repositioning effect JON, accepted JON, accepted • Better grey matter DTI due to reduced FA MacNab et al., contrast inhomogeneity ISMRM, 2011

Overview • Diffusion Tensor Imaging (DTI) in brief • Example application in DTI • Three artefacts in DTI – Eddy Current (EC) distortions – Local Perturbation Fields (LPFs) – Signal -dropout due to mechanical vibration • Take home message

Vibration artefacts in blip up and blip down DTI data sets Gallichan et al., HBM, 2010

Problem: signal -dropout due to axial rotation Unshifted echo (blip-up PE) [arbitrary units] k-space coverage echo 1 k y /PE 0 k min k max k =0 Shifted echo (blip-up PE) [arbitrary units] ∆ k } y ( ) 1 ∆ ∝ − Ω k m eff r 1 y z k y /PE 0 Mohammadi et al., MRM, accepted k min k max k =0

Recover signal using phase encoding reversal Blip down Blip up Mohammadi et al., MRM, accepted

Correction of vibration artefacts in DTI using phase-encoding reversal (COVIPER) ed Mohammadi et al., MRM, accepted

Relevance • Robust data, e.g., avoiding false positives in FA group studies Keller et al., JON, accepted • Better data quality in grey matter MacNab et al., ISMRM, 2011 • Less signal -dropout artefacts in HARDI Aganj et al., MRM, 2010

Take home message • Retrospective artefact correction is possible • Sensitivity and robustness of DTI can be improved • • Three artefacts related to the diffusion weighting gradients were Three artefacts related to the diffusion weighting gradients were presented • We are not finished yet

Acknowledgements • MR physics group in WTCN, London – Nikolaus Weiskopf (my supervisor and head of MR physics at the WTCN) – Zoltan Nagy – Oliver Josephs Chloe Hutton (special thanks for the acronym ☺ ) – – Antoine Lutti • • External collaborators External collaborators – Michael Deppe (University of Münster) – Harald Möller (Max Plank Institute Leipzig) – Dirk Müller (University of Münster) – Mark Symms (Department of Clinical and Experimental Epilepsy, UCL, London) – David Carmichael (Imaging and Biophysics, UCL, London) This work was supported by the Wellcome Trust.