Vincent Kyu Lee1,2, Rafael Ceschin3,4, William Thomas Reynolds2,3, Benjamin Meyers4, Julia Wallace4, Douglas Landsittel5, Daryaneh Badaly6, J William Gaynor7, Daniel Licht8, Nathaniel H Greene9, Ken M Brady10, Jill V Hunter11, Zili D Chu12, Elisabeth A Wilde13, R Blaine Easley14, Dean Andropoulos14, and Ashok Panigrahy1,2,3,4
1Bioengineering, University of Pittsburgh, Pittsburgh, PA, United States, 2Radiology, UPMC Children's Hospital of Pittsburgh, Pittsburgh, PA, United States, 3Biomedical Informatics, University of Pittsburgh, Pittsburgh, PA, United States, 4Radiology, University of Pittsburgh, Pittsburgh, PA, United States, 5Epidemiology and Biostatistics, Indiana University School of Public Health-Bloomington, Bloomington, IN, United States, 6Learning and Development Center, Child Mind Institute, New York, NY, United States, 7Division of Cardiothoracic Surgery, Children’s Hospital of Philadelphia, Philadelphia, PA, United States, 8Division of Neurology, Children’s Hospital of Philadelphia, Philadelphia, PA, United States, 9Anesthesiology, Oregon Health and Sciences University, Portland, OR, United States, 10Anesthesiology, Lurie Children’s Hospital, Northwestern University, Chicago, IL, United States, 11Radiology, Texas Children's Hospital, Houston, TX, United States, 12Radiology, Texas Children's Hospital, Baylor College of Medicine, Houston, TX, United States, 13Neurology, University of Utah School of Medicine, Salt Lake City, UT, United States, 14Anesthesiology, Perioperative and Pain Medicine, Texas Children's Hospital, Baylor College of Medicine, Houston, TX, United States
Synopsis
This study
examined trajectories of early postnatal brain structure (macrostructural brain
volumes and microstructural white matter tractography) relationship with early
childhood neurodevelopmental deficits (NDD) in complex congenital heart disease
patients. This analysis included development of predictive multi-variable
models incorporating other known risk factors of poor NDD in CHD. A multi-modal
dysmaturation phenotype of reduced subcortical volume and cerebral white matter
volume/connectivity predicted poor early childhood language outcomes,
despite high relative contribution of genetic and socio-demographic factors
(maternal IQ). Favorable socio-demographic factors, despite the high incidence
of focal WMI and presence of genetic abnormalities, predicted better
neurodevelopmental outcomes.
INTRODUCTION
Congenital heart defects (CHD), especially the severe forms – such as hypoplastic left heart syndrome (HLHS) and transposition of the great arteries (TGA) – have been shown to adversely affect neurodevelopment1,2. These neurodevelopmental deficits (NDD) include impairments in gross cognitive/intellectual functions, language/verbal abilities, and visual and motor skills in children with CHD1,3. Our primary objective was to correlate trajectories of early postnatal brain structure with early childhood NDD in complex CHD patients. Multi-modal early serial postnatal magnetic resonance imaging (MRI) – macrostructure (regional brain volumes) and microstructure fractional anisotropy (FA) of white matter (WM) tracts – was performed. Our secondary objectives were to develop predictive multi-variable models incorporating known risk factors of poor NDD in CHD.METHODS
A total of 99
patients (53 male) with HLHS or TGA were prospectively recruited at a single
site (Texas Children’s Hospital) and received three MRI scans – preoperatively (39.42±1.53
weeks post-conceptual age at scan), postoperatively #1 (40.66±1.53 weeks), and postoperatively
#2 (61.34±7.51 weeks). All MRI scans were acquired on the same 1.5T Philips
Intera scanner (Philips Medical Systems, Best, Netherlands). Each of the scans
included volumetric 3D T1 imaging (TE/TR=4.0629/20ms, FOV=256mm, voxel size=1mmx1mmx1mm,
flip angle=30°) and diffusion imaging (TE/TR=90/6065ms, FOV=200mm, voxel size=2mmx2mmx2.7mm,
15 gradient directions B=860s/mm2). All volumetric T1 were processed
through an in-house semi-automated segmentation pipeline, Neonatal Brain
Structure Segmentation4,5; and segmented into
following regions: brainstem, cerebellum, cortex, intracranial cerebrospinal
fluid (CSF), deep grey matter (GM), WM, whole brain (excluding CSF). Each DTI
acquisition were reconstructed using diffusion tensor algorithm in DSI Studio6.
Focal white matter injury (WMI) (high T1 signal) was assessed both qualitatively
by neuroradiologist review (presence/absence) and continuously (volume by Licht
et al). The processed DTI white matter tracts included the following: genu,
body, and splenium of the corpus callosum; left and right cortico-spinal tract
(CSTL and CSTR), fronto-occipital fasciculus (FOFL and FOFR), inferior
longitudinal fasciculus (ILFL and ILFR), and superior longitudinal fasciculus
(SLFL and SLFR). Cognitive, language, and visual/motor outcomes were assessed
with the Bayley-III at 1 and 3 years. At 5 years, the full-scale IQ and verbal
IQ of the WPPSI-III and the Beery VMI were used (as indices of similar outcomes
to the three indices on the Bayley-III).
Our three primary multi-variable models incorporated the longest epoch or overall imaging
trajectory (preoperative to follow-up) predicting 5-year outcomes (n=34);
3-year outcomes (n=35); 1-year outcomes (n=45). Our secondary models
incorporated two epochs of earlier trajectory imaging measures: perioperative
trajectory (preoperative to postoperative) predicting 5-year outcomes (n=36);
3-year outcomes (n=35); 1-year outcomes (n=49); and post-surgical trajectory (postoperative
#1 to postoperative #2) predicting 5-year outcomes (n=35); 3-year outcomes (n=37);
1-year outcomes (n=44). Single ventricle, maternal IQ and genetic abnormality
were used as co-variates in the final model as they predicted a greater
degree/range of ND outcome variance compared to other factors listed in Table 1.
Lastly, a post-hoc factor contribution analysis was conducted for each ND test
to examine the contributions of each independent variable to the variance in
outcomes (Figures 1-3).RESULTS
Cohort
characteristics are presented in Table 1.
Overall
Period Multi-modal Trajectory and Early Childhood Outcome: Decreased trajectory of brainstem
(p=0.0022) and WM volume (p=0.0397) predicted poor 5-year verbal IQ, while
decreased trajectory of brainstem/deep GM volume and decreased trajectory of
fractional anisotropy (FA) of SLFR (p=0.0256) predicted poor 3-year language
outcome independent of maternal IQ/genetic abnormality.
Overall
Period Multi-modal Trajectory and Infant Outcomes: Decreased trajectory of whole brain
volume (p=0.0286) predicted poor 1-year language outcome independent of
maternal IQ/genetic abnormality.
Post-surgical
Microstructure (FA) Trajectory and Early Childhood Outcomes: Decreased trajectory FA of the ILFL (p=0.0354), ILFR
(p=0.0156) predicted poor 3-year motor outcomes while decreased trajectory of
FA of FOFR (p=0.0082) predicted poor 3-year cognitive outcomes independent of lower
maternal IQ/presence of genetic abnormality/single ventricle status.
Perioperative
Multi-modal Trajectory and Early Childhood Outcomes: Decreased trajectory of brainstem
volume (p=0.0111) and reduced trajectory of FA of SLFL (p=0.0090) predicted
poor 3-year language outcomes independent of maternal IQ.
Perioperative
Multi-modal Trajectory and Infant Outcomes: Increased CSF volume (p=0.0062) and decreased trajectory of FA of ILFL (p=0.0402), ILFR (p=0.0112), FOFL (p=0.0402), FOFR
(p=0.0089), SLFL (p=0.0042) predicted poor 1-year cognitive outcomes
independent of maternal IQ/genetic abnormality.DISCUSSION
Increased trajectory of CSF volume between the pre-operative and
post-operative scan contributed to the greatest degree of infant ND variance
(64%) relative to all imaging metrics. Post-surgical trajectories of
microstructure (DTI), but not macrostructure (volume), were predictive of early
childhood (3-year) cognitive and motor outcomes, suggesting vulnerability of
cerebral white matter connectivity in the post-surgical period. Focal white
matter injury qualitative and quantitative measures did not predict ND outcomes.
A multi-modal dysmaturation phenotype of reduced subcortical volume (deep GM,
brainstem) and cerebral white matter volume/connectivity predicted poor early
childhood language outcomes, despite high relative contribution of genetic and
socio-demographic factors (maternal IQ). Favorable socio-demographic factors
(high maternal IQ, high SES), despite the presence of a high incidence of focal
WMI and genetic abnormalities, were predictive of better ND outcomes suggesting
the possibility of the presence of reserve mechanisms in this cohort.CONCLUSION
Reduced trajectory of multi-modal infant brain maturation (reduced
subcortical volume and long-range cortico-cortical connectivity) predicts poor
early childhood language outcomes independent of genetic and socio-demographic factors. Acknowledgements
Christine Johnson, Nancy H. Beluk, Jennifer Frye, Tasha Hwostow.References
- Newburger, J.W., et al., Early developmental outcome in children with hypoplastic left heart syndrome and related anomalies: the single ventricle reconstruction trial. Circulation, 2012. 125(17): p. 2081-2091.
- Bellinger, D. and J. Newburger, A Longitudinal Study From Infancy to Adolescence of the Neurodevelopmental Phenotype Associated With d-Transposition of the Great Arteries, in Congenital Heart Disease and Neurodevelopment. 2016, Elsevier. p. 27-40.
- Marelli, A., et al., Brain in congenital heart disease across the lifespan: the cumulative burden of injury. Circulation, 2016. 133(20): p. 1951-1962.
- Ceschin, R., et al., A computational framework for the detection of subcortical brain dysmaturation in neonatal MRI using 3D Convolutional Neural Networks. NeuroImage, 2018. 178: p. 183-197.
- Ceschin, R. (2021, September 20). NeBSS (Version 2021 September). Github. https://github.com/PIRCImagingTools/NeBSS
- Yeh, F. (2021, May 15). DSI Studio (Version 2021 May). Zenodo. http://doi.org/10.5281/zenodo.4764264