TY - GEN
T1 - Compositional measures of diffusion anisotropy and asymmetry
AU - Cieslak, Matthew
AU - Meiring, Wendy
AU - Brennan, Tegan
AU - Greene, Clint
AU - Volz, Lukas J.
AU - Vettel, Jean M.
AU - Suri, Subhash
AU - Grafton, Scott T.
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/5/23
Y1 - 2018/5/23
N2 - Diffusion MRI (dMRI) can be used to examine white matter structures in the living human brain. A common approach to dMRI analysis involves calculating a scalar value that reflects how different the observed diffusion is from an isotropic diffusion profile in each voxel. Current methods require either an oversimplified model (Fractional anisotropy, FA), use an abstract measure of 'sharpness' (Generalized Fractional Anisotropy, GFA) or perform inconsistently across scans (Quantitative anisotropy, QA). Here we propose two novel statistically-motivated anisotropy measures based on fiber transition probabilities to neighboring voxels, calculated from orientation distribution functions (ODFs) using recently developed analytic tractography. Compositional Distance from Isotropy (CoDI) is the Aitchison distance between a voxel's analytic transition probabilities to its 26 neighbors and the corresponding transition probabilities that would arise from an isotropic ODF. Compositional Asymmetry (CoAsy) incorporates ODF similarities between neighboring voxels to highlight asymmetric patterns in transition probabilities to better reflect complexities of white matter structures. CoAsy is higher where fascicles split/fan/curve within white matter. We demonstrate on a fiber phantom that CoAsy reflects these properties. Using a cohort of 25 individuals, each scanned 8 times, we show that CoDI and CoAsy values reflect underlying fiber populations and that these measurements are reproducible across repeated scans.
AB - Diffusion MRI (dMRI) can be used to examine white matter structures in the living human brain. A common approach to dMRI analysis involves calculating a scalar value that reflects how different the observed diffusion is from an isotropic diffusion profile in each voxel. Current methods require either an oversimplified model (Fractional anisotropy, FA), use an abstract measure of 'sharpness' (Generalized Fractional Anisotropy, GFA) or perform inconsistently across scans (Quantitative anisotropy, QA). Here we propose two novel statistically-motivated anisotropy measures based on fiber transition probabilities to neighboring voxels, calculated from orientation distribution functions (ODFs) using recently developed analytic tractography. Compositional Distance from Isotropy (CoDI) is the Aitchison distance between a voxel's analytic transition probabilities to its 26 neighbors and the corresponding transition probabilities that would arise from an isotropic ODF. Compositional Asymmetry (CoAsy) incorporates ODF similarities between neighboring voxels to highlight asymmetric patterns in transition probabilities to better reflect complexities of white matter structures. CoAsy is higher where fascicles split/fan/curve within white matter. We demonstrate on a fiber phantom that CoAsy reflects these properties. Using a cohort of 25 individuals, each scanned 8 times, we show that CoDI and CoAsy values reflect underlying fiber populations and that these measurements are reproducible across repeated scans.
KW - Anisotropy
KW - Compositional Data Analysis
KW - Diffusion MRI
KW - Segmentation
UR - http://www.scopus.com/inward/record.url?scp=85048119092&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85048119092&partnerID=8YFLogxK
U2 - 10.1109/ISBI.2018.8363537
DO - 10.1109/ISBI.2018.8363537
M3 - Conference contribution
AN - SCOPUS:85048119092
T3 - Proceedings - International Symposium on Biomedical Imaging
SP - 123
EP - 126
BT - 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018
T2 - 15th IEEE International Symposium on Biomedical Imaging, ISBI 2018
Y2 - 4 April 2018 through 7 April 2018
ER -