Youngseob Seo PhD, Presenter: Nothing to Disclose

Zhiyue Jerry Wang PhD, Abstract Co-Author: Nothing to Disclose

Jonathan Chia, Abstract Co-Author: Employee, Koninklijke Philips Electronics NV

Michael Craig Morriss MD, Abstract Co-Author: Nothing to Disclose

Nancy Katherine Rollins MD, Abstract Co-Author: Nothing to Disclose

Clinical diffusion tensor imaging (DTI) is usually acquired with parallel imaging which can lower SNR non-uniformly across the brain. Low SNR can bias fractional anisotrophy (FA). We determined the thresholds below which low SNR caused a measurable bias in FA and also compared SNR and FA derived using manual ROI analysis to that derived using Tract-Based Spatial Statistics (TBSS), an operator-independent whole brain analysis tool.

A normal adult male served as the phantom. DTI was performed at 1.5T and 3T using SS-EPI; 25.6cm2 FOV; 2 mm3 voxels; 42 slices; TR/TE=6218/100(1.5T) and 4344/74(3T); b=700; SENSE=2; 30 directions; and 15 acquisitions. Affine transformation, eddy current correction and signal averaging were done using PRIDE. FA and SNR were measured after 10 incremental additions of Gaussian noise to the diffusion data set derived using 15 acquisitions; noise was assessed from the subtraction of two magnitude images of consecutive acquisitions; SNR was calculated from b=0 images using IDL. FA was measured on each of the 10 DTI data sets using manual ROI analysis and TBSS. A single observer manually placed ROIs on high FA white matter (WM), e.g. callosal genu (CG), intermediate FA WM, e.g. subcortical U fiber, and low FA structure, e.g. putamen (PUT). For TBSS, all FA images were aligned using nonlinear registration and affine transformed into MNI152 space. Mean FA image was used to create an FA skeleton (FA threshold=0.2); the the CG and U fiber were studied using TBSS; FA<0.2 of the PUT precluded analysis with TBSS.

ROI analysis at 1.5T and 3T showed that low SNR caused a downward bias in regions of high FA (CG) and upward bias in regions of intermediate and low FA (U fiber and PUT). However, TBSS showed an upward bias in the CG and U fiber at 1.5T and 3T. Using ROI analysis, SNR thresholds for bias in FA were 10(1.5T) and 20(3T) in the CG, 30(1.5T) and 50(3T) in the U fibers, and 55(1.5T) and 80(3T) in the PUT. Using TBSS, SNR threshold was 20 at 1.5T and 3T in the CG, 25(1.5T) and 35(3T) in the U fiber.

SNR thresholds below which FA is biased vary with the analytical methodolgy, region of the brain, and field strength while the direction of the bias in FA at low SNR differs between ROI analysis and TBSS.

For reproducible assessment of FA, there should be documentation that SNR is above critical thresholds.

Seo, Y, Wang, Z, Chia, J, Morriss, M, Rollins, N, Determination of Lower Limits of SNR in Diffusion Tensor Imaging: A Comparison of Two Techniques.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9013035.html