TY - CHAP
T1 - Regression analysis for assessment of myelination status in preterm brains with magnetic resonance imaging
AU - Wang, Siying
AU - Murgasova, Maria
AU - Hajnal, Joseph Vilmos
AU - Ledig, Christian
AU - Schnabel, Julia Anne
PY - 2016/4
Y1 - 2016/4
N2 - Myelination is considered an important developmental process during human brain maturation and to be closely correlated with gestational age. Assessment of the myelination status generally requires dedicated imaging, yet the conventional T2-weighted acquisitions routinely obtained during clinical imaging of neonates carry signatures that are thought to be directly associated with myelination. In this work, we propose a method to identify these signatures which could potentially be used to assess brain maturation of preterm neonates directly from T2-weighted magnetic resonance images. First we segment the tissue that is likely to contain myelin from 96 preterm neonates. We then construct a spatio-temporal atlas based on the registered segmentations by fitting a voxelwise logistic regression model. Finally, the atlas is utilized to estimate the gestational ages of individual subjects in a leave-one-out procedure. The logistic model yields a root mean squared error of 10 days, as compared to 13 days for the ages predicted using a kernel regression atlas.
AB - Myelination is considered an important developmental process during human brain maturation and to be closely correlated with gestational age. Assessment of the myelination status generally requires dedicated imaging, yet the conventional T2-weighted acquisitions routinely obtained during clinical imaging of neonates carry signatures that are thought to be directly associated with myelination. In this work, we propose a method to identify these signatures which could potentially be used to assess brain maturation of preterm neonates directly from T2-weighted magnetic resonance images. First we segment the tissue that is likely to contain myelin from 96 preterm neonates. We then construct a spatio-temporal atlas based on the registered segmentations by fitting a voxelwise logistic regression model. Finally, the atlas is utilized to estimate the gestational ages of individual subjects in a leave-one-out procedure. The logistic model yields a root mean squared error of 10 days, as compared to 13 days for the ages predicted using a kernel regression atlas.
U2 - 10.1109/ISBI.2016.7493263
DO - 10.1109/ISBI.2016.7493263
M3 - Conference paper
BT - 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI)
PB - IEEE
ER -