Methodological improvements in voxel‐based analysis of diffusion tensor images: Applications to study the impact of apolipoprotein E on white matter integrity

Abstract

Purpose To identify regional differences in apparent diffusion coefficient (ADC) and fractional anisotropy (FA) using customized preprocessing prior to voxel-based analysis (VBA) in 14 normal subjects with the specific genes that decrease (APOE ε2) and that increase (APOE ε4) the risk of Alzheimer’s Disease. Materials and Methods Diffusion Tensor images (DTI) acquired at 1.5 T were denoised with a total variation tensor regularization algorithm prior to affine and nonlinear registration to generate a common reference frame for the image volumes of all subjects. Anisotropic and isotropic smoothing with varying kernel sizes was applied to the aligned data prior to VBA to determine regional differences between cohorts segregated by allele status. Results VBA on the denoised tensor data identified regions of reduced FA in APOE ε4 compared to the APOE ε2 healthy older carriers. The most consistent results were obtained using the denoised tensor and anisotropic smoothing prior to statistical testing. In contrast, isotropic smoothing identified regional differences for small filter sizes alone, emphasizing that this method introduces bias in FA values for higher kernel sizes. Conclusion Voxel-based DTI analysis can be performed on low SNR images to detect subtle regional differences in cohorts using the proposed pre-processing techniques.

Imaging Alzheimer's genetic risk using diffusion MRI: A systematic review

“… • Carriers vs. non-carriers: ↓ FA in right ventral cingulum & left inferior longitudinal fasciculus Ma et al, 2017 885 (145; 729) 65.3 (7.4) Structural connectivity 3 T 1000 Not reported 30 1 Whole brain structural network •Healthy carriers vs healthy non-carriers: ↑ clustering coefficient & local efficiency • MCI carriers vs. non-carriers: ↓ clustering coefficient & local efficiency • All carriers vs. non-carriers: ↓ nodal efficiency in: inferior frontal gyrus, orbital part; left superior frontal gyrus, orbital part; left middle occipital gyrus. ↑ nodal efficiency in: left cuneus; left inferior parietal but supramarginal and angular gyri Newlander et al, 2014 14 (7; 7) 72.7 (6.1) VBA 1.5 0, 800 1 × 1 × 1 12 1 Whole brain • Carriers vs. non-carriers: ↓ FA in genu of corpus callosum & brain stem Nierenberg et al, 2005 29 (14; 15) 67.1 (6.5) ROI 1.5 0, 1000 Not reported 20 1 Parahippocampal cingulum • Carriers vs. non-carriers: ↓ FA & ↑ RD in parahippocampal cingulum Nyberg et al, 2014 273 (69; 204) 67.01 (8.0) TBSS, VBA 3 1000 0.98 × 0.98 × 2 32 1 Whole brain • Carriers vs. controls: no significant difference for whole brain metrics or specific subregion metrics in TBSS. ↓ FA in five anterior and posterior midline regions on VBA O'Dwyer et al, 2012 44 (22; 22) 26.7 (4.0) TBSS 3 1000 2 × 2 × 2 60 1 Whole brain • Carrier vs. non-carrier: no significant differences in diffusion indices • Carrier/non-carrier prediction accuracy: sensitivity & specificity range 93–100% using a feature selection algorithm, support vector machines & FA data Operto et al, 2018 532 (275; 257) 58.1 (7.5) TBSS 3 ...…”

“…Reduced neurite density index (NDI) and increased free isotropic water fraction (FISO) are also reported. The white matter regions found to be associated with APOE status included: the genu ( Newlander et al, 2014 , Zhang et al, 2015 , Cai et al, 2017 , Cavedo et al, 2017 ), body ( Persson et al, 2006 , Zhang et al, 2015 ) and splenium ( Ryan et al, 2011 , Slattery et al, 2017 ) of the corpus callosum and the corpus callosum overall ( Heise et al, 2011 , Westlye et al, 2012 , Cavedo et al, 2017 ); the parahippocampal cingulum ( Nierenberg et al, 2005 , Bagepally et al, 2012a , Bagepally et al, 2012b , Kljajevic et al, 2014 , Zhang et al, 2015 ) and the cingulum overall ( Adluru et al, 2014 , Lyall et al, 2014 , Cavedo et al, 2017 ); the intracalacrine sulcus ( Bagepally et al, 2012a , Bagepally et al, 2012b , Westlye et al, 2012 ); the brain stem ( Westlye et al, 2012 , Newlander et al, 2014 ); the corona radiata ( Heise et al, 2011 , Smith et al, 2016 , Cai et al, 2017 , Cavedo et al, 2017 , Slattery et al, 2017 , Operto et al, 2018 ); the external capsule ( Heise et al, 2011 , Cavedo et al, 2017 ) and internal capsule ( Heise et al, 2011 , Westlye et al, 2012 , Smith et al, 2016 , Cavedo et al, 2017 ); the superior longitudinal fasciculus ( Adluru et al, 2014 , Lyall et al, 2014 , Cavedo et al, 2017 , Operto et al, 2018 ) and inferior longitudinal fasciculus ( Dowell et al, 2013 , Cavedo et al, 2017 ); the fronto-occipital fasiculus ( Cavedo et al, 2017 , Operto et al, 2018 ); the fornix ( Zhang et al, 2015 ); the cerebral peducles ( Zhang et al, 2015 ); the cortico-spinal tract ( Lauk...…”

“… Tract Diagram Summary of APOE Findings Summary of FAD Findings Corpus Callosum: Connects the left and right cerebral hemispheres E4 carriers vs . non-carriers: ↓FA ( Persson et al, 2006 , Newlander et al, 2014 , Heise et al, 2010, Cavedo et al, 2017 , Zhang et al, 2015 , Ryan et al, 2011 , Slattery et al, 2017 ) ‘ ↑MD (Heise et al, 2010, Cavedo et al, 2017 , Cai et al, 2017 , Zhang et al, 2015 ) ↑RD ( Cavedo et al, 2017 , Westlye et al, 2012 ) Pre-symptomatic carriers vs . controls: ↓FA ( Sanchez-Valle et al, 2016 ) Symptomatic carriers vs. controls: ↑MD ( Sanchez-Valle et al, 2016 , Parra et al, 2015 ) ↑LD ( Sanchez-Valle et al, 2016 ) ↑RD ( Sanchez-Valle et al, 2016 ) Cingulum: Connects the temporal and frontal lobes, cingulate and medial gyri of frontal, occipital, parietal and temporal lobes E4 carriers vs .…”

Primary Blast Traumatic Brain Injury in the Rat: Relating Diffusion Tensor Imaging and Behavior

“…An iterative procedure using rigid body, affine, and diffeomorphic registration was used to create a template at a final resolution of 200 μm 3 . Tensor volumes were spatially smoothed with an anisotropic three-dimensional filter implemented in Matlab (39). Maps of fractional anisotropy (FA) were derived and used for subsequent statistical analysis.…”

Fully Automated and Adaptive Intensity Normalization Using Statistical Features for Brain MR Images

“…In the literature, intensity normalization is mostly performed by Gaussian filtering [10][11][12][13][14]39]. However, results obtained from an original image ( Figure 3.a) by this method may not be efficient in case of a chosen small σ value (Figure 3.b).…”