TY - JOUR
T1 - More insights into early brain development through statistical analyses of eigen-structural elements of diffusion tensor imaging using multivariate adaptive regression splines
AU - Chen, Yasheng
AU - Zhu, Hongtu
AU - An, Hongyu
AU - Armao, Diane
AU - Shen, Dinggang
AU - Gilmore, John H.
AU - Lin, Weili
N1 - Funding Information:
This study was supported in part by NSF Grant BCS-08-26844 and NIH Grants RR025747-01, P01CA142538-01, MH086633, and AG033387, NIH Grants 1R01EB006733, R01EB008374, and 1R01EB009634, NIH Grants R01MH070890 and R01HD053000, and NIH Grant R01NS055754.
PY - 2014/3
Y1 - 2014/3
N2 - The aim of this study was to characterize the maturational changes of the three eigenvalues (λ1 ≥ λ2 ≥ λ3) of diffusion tensor imaging (DTI) during early postnatal life for more insights into early brain development. In order to overcome the limitations of using presumed growth trajectories for regression analysis, we employed Multivariate Adaptive Regression Splines (MARS) to derive data-driven growth trajectories for the three eigenvalues. We further employed Generalized Estimating Equations (GEE) to carry out statistical inferences on the growth trajectories obtained with MARS. With a total of 71 longitudinal datasets acquired from 29 healthy, full-term pediatric subjects, we found that the growth velocities of the three eigenvalues were highly correlated, but significantly different from each other. This paradox suggested the existence of mechanisms coordinating the maturations of the three eigenvalues even though different physiological origins may be responsible for their temporal evolutions. Furthermore, our results revealed the limitations of using the average of λ2 and λ3 as the radial diffusivity in interpreting DTI findings during early brain development because these two eigenvalues had significantly different growth velocities even in central white matter. In addition, based upon the three eigenvalues, we have documented the growth trajectory differences between central and peripheral white matter, between anterior and posterior limbs of internal capsule, and between inferior and superior longitudinal fasciculus. Taken together, we have demonstrated that more insights into early brain maturation can be gained through analyzing eigen-structural elements of DTI.
AB - The aim of this study was to characterize the maturational changes of the three eigenvalues (λ1 ≥ λ2 ≥ λ3) of diffusion tensor imaging (DTI) during early postnatal life for more insights into early brain development. In order to overcome the limitations of using presumed growth trajectories for regression analysis, we employed Multivariate Adaptive Regression Splines (MARS) to derive data-driven growth trajectories for the three eigenvalues. We further employed Generalized Estimating Equations (GEE) to carry out statistical inferences on the growth trajectories obtained with MARS. With a total of 71 longitudinal datasets acquired from 29 healthy, full-term pediatric subjects, we found that the growth velocities of the three eigenvalues were highly correlated, but significantly different from each other. This paradox suggested the existence of mechanisms coordinating the maturations of the three eigenvalues even though different physiological origins may be responsible for their temporal evolutions. Furthermore, our results revealed the limitations of using the average of λ2 and λ3 as the radial diffusivity in interpreting DTI findings during early brain development because these two eigenvalues had significantly different growth velocities even in central white matter. In addition, based upon the three eigenvalues, we have documented the growth trajectory differences between central and peripheral white matter, between anterior and posterior limbs of internal capsule, and between inferior and superior longitudinal fasciculus. Taken together, we have demonstrated that more insights into early brain maturation can be gained through analyzing eigen-structural elements of DTI.
KW - DTI longitudinal analysis
KW - DTI regression analysis
KW - Early brain development
KW - GEE
KW - Multivariate adaptive regression splines
UR - http://www.scopus.com/inward/record.url?scp=84896047701&partnerID=8YFLogxK
U2 - 10.1007/s00429-013-0517-7
DO - 10.1007/s00429-013-0517-7
M3 - Article
C2 - 23455648
AN - SCOPUS:84896047701
SN - 1863-2653
VL - 219
SP - 551
EP - 569
JO - Brain Structure and Function
JF - Brain Structure and Function
IS - 2
ER -