CNI Inauguration Workshop Stanford, March 22 nd, 2012 High Fidelity Brain Connectivity Imaging -Recent Progress on Diffusion Weighted MRI for High Resolution and Low Distortion Allen W. Song, PhD Brain Imaging and Analysis Center Duke University
The Need for High Resolution and Low Distortion to Image Connectivity Brain as an Example: >100 Billion Neurons >1 Trillion Connections >7 Billions Brains
What Resolution Should We Reach with Diffusion Weighted MRI? High Angular Resolution Solves Crossing Fibers without Demand on Spatial Resolution High Spatial Resolution Is Needed for U-fibersU
What Resolution Should We Reach with Diffusion Weighted MRI? High-Resolution Susceptibility Imaging Guidon,, Liu, et al., Duke diamagnetic 0 paramagnetic WM/GM 12 34 mm into Interface GM WM
Benefit of High-Resolution Diffusion MRI High-Res FA 1 mm 3 ROIs 1x1x1 mm 3 2x2x2 mm 3 3x3x3 mm 3 2x2x2 mm 3 1x1x1 mm 3 High-Resolution DTI Liu et al., Duke
Benefit of High-Resolution Diffusion MRI Region 1 1 234 0.8 FA 234 3x3x3mm 3 2x2x2mm 3 1x1x1mm 3 0 High-Resolution DTI Guidon et al., Duke
High Resolution DWI Is Not So Easy. At the present time, the only viable solution to achieve high-resolution DWI is through multi-shot acquisition Two-shot DW EPI Two-shot DW Spiral
Self-Navigated Interleaved Spiral (SNAILS) To Remove Aliasing Artifacts bo b1 b2 Baseline Image Diffusion Weighting Images But it requires specialized pulse sequence, as well as additional scan and reconstruction time Liu et al., BIAC, Duke University
Goal: To develop efficient, effective, and practical acquisition/reconstruction methods for high- fidelity DWI. Criteria: Ideally, the methodology should require minimal pulse sequence changes, remove the need for additional reference scans, and still result in high-quality DW images.
Cycling-Phase Reconstruction (CPR) A general post-processing processing framework designed to correct for hardware- and subject-induced aliasing artifacts in accelerated MRI scans (EPI, Spiral, Propeller, FSE, GRASE among others), without reference scans Also applicable to diffusion weighted single-shot shot / multi-shot / parallel EPI, Spiral, and Propeller acquisitions, without reference scans
CPR for 2D EPI Intensity Profile Sorted Profile Cost Function Chen et al.,magn Reson Med. 2011 Oct;66(4):1057-66
CPR for 2D EPI Chen et al.,magn Reson Med. 2011 Oct;66(4):1057-66
No Correction 1-D D Reference CPR 1-Shot 2-Shot 3-Shot 4-Shot EPI Chen et al.,magn Reson Med. 2011 Oct;66(4):1057-66
Can CPR Work in Spiral Imaging? a: b=0 image b: image (a) with simulated phase error d: background energy in images (c) as a function of φ 0 and g x c: image (b) reconstructed by cycling phases with different φ 0 and g x values the image with the minimum background energy is the one with the least aliasing Truong et al. (in press) Magn. Reson. Med.
CPR for Two-Shot DW Spiral Imaging shot 1 aliased images k-space trajectory point spread function (N x 0 +1:2N x 0, N y 0 +1:2N y 0 ) motioninduced phase error known unaliased image unknown shot 2 a 1 a 2 P 11 = Eq. [1] e iφ P 2N 2 e iφ P 21 P 1N 2 u 1 u n u N 2 The 2 nd row of E contains the motion-induced phase term exp[ iφ(x 0, y 0 )]
CPR for Two-Shot DW Spiral Imaging φ is assumed to be spatially linear: φ(x, y) = φ 0 + xg x + yg y which is sufficient to correct for rigid-body motion* A series of images are reconstructed by solving Eq. [1] with different φ 0, g x, and g y values S background φ 0 / g x / g y min The final image is chosen as the one with the lowest signal intensity in the background, i.e., with the least amount of aliasing *Anderson & Gore (1994) MRM 32: 379 87
CPR Improves Diffusion Weighted Multi-Shot High-Resolution Spiral Imaging uncorrected DWI images corrected DWI images uncorrected ADC maps corrected ADC maps Truong et al. (in press) Magn. Reson. Med.
CPR Improves Diffusion Weighted Multi-Shot EPI After Correction Improvement in two-shot DW-EPI
CPR Improves Four-Shot DW EPI Original Images with Reference Echoes Original Image with Reference Echoes Improved Image after CPR
Geometric Distortion is Another Major Challenge for High-Resolution DWI Distortion usually proportional to the readout window length 160ms 128 echoes 2.5 mm 3 320mm 100ms 64 echoes SENSE Acceleration Factor 2 2.5 mm 3 320mm
EPI Distortions from Static Bo Inhomogeneity
DW EPI Distortions from Dynamic Eddy Current DW EPI DW Spiral
Mapping Eddy Current Induced Field Changes Truong et al., NeuroImage. 2011 Aug 15;57(4):1343-7
Field Map Based Distortion Correction Truong et al., NeuroImage. 2011 Aug 15;57(4):1343-7
Field Map Based Deblurring in Spiral Imaging [0.00, 0.00, 0.00] [0.79, 0.32, 0.51] [ 0.66, 0.16, 0.72] bo Diffusion Direction #1 Diffusion Direction #2 Avram et al., ISMRM 2011
Can We Make Field Map Generation Inherent? Echo-shifting effect in EPI Chen et al., Neuroimage. 2006; 31(2):609-22 Truong et al., Magn. Reson. Med. 2010; 64:1121-7
Inherent Field Mapping based on K-space Energy Spectrum Analysis (KESA) Chen et al., Neuroimage. 2006; 31(2):609-22 Truong et al., Magn. Reson. Med. 2010; 64:1121-7
Echo-Shifting Effect in DTI
Magnetic Field Maps in DTI
Summary A practical yet powerful technical foundation for improved DWI / DTI acquisition and reconstruction strategies Inherent correction for ghosting artifacts using cycling phase reconstruction (CPR) without the need for potentially time-consuming reference echoes in high-resolution DWI Dynamic distortion correction based on inherently generated field maps to restore spatial fidelity Generally applicable to all fast imaging techniques including EPI, Spiral, Propeller, and other k-space k trajectories
Acknowledgement Nan-Kuei Chen Alex Avram Chunlei Liu Arnaud Guidon Todd Harshbarger Ana Batrachenko Trong-Kha Truong NS 041328, EB 009483, NS 075017 Faculty and Staff at the Duke UNC Brain Imaging and Analysis Center at Duke University www.biac.duke.edu