This investigation included data acquired from an Autism Center of Excellence network study funded by the National Institutes of Health. Informally called the Infant Brain Imaging Study (IBIS), the network includes four clinical data collection sites (University of North Carolina at Chapel Hill, University of Washington, Children's Hospital of Philadelphia, Washington University in St Louis), a data coordinating center (Montreal Neurological Institute at McGill University), and two image processing sites (University of Utah, University of North Carolina). The network study collects neuroimaging and behavioral data on infants who are at genetic risk for ASDs by virtue of having an older sibling with autism, referred to as “high-risk infant siblings.” The study enrolls 1) 6-month-old high-risk infant siblings who are seen for follow-up assessments at 12 and 24 months of age, 2) high-risk infant siblings who enter the study at 12 months of age and are followed up at 24 months, and 3) comparison subjects who are typically developing infants considered to be at low risk for autism (i.e., an older sibling is developing typically), who are seen at 6, 12 and 24 months. Diagnostic outcome data for ASD status is determined at 24 months of age (and again at 36 months of age if the timeline of the project allows). For the purposes of this article, we have included cross-sectional data on the 6-month-olds who were scanned from the start of the study through the fall of 2010. Two groups were included: infants at high risk for autism and typically developing children considered to be at low risk for autism. Subjects were combined from all four clinical data collections sites.
The subjects include all children with imaging data collected and processed through the end of September 2010. There were 318 children who attempted the MRI scans, and 88% of the high-risk infants and 83% of the low-risk subjects were successfully scanned, for a total of 276 scans. Owing to difficulties in scheduling visits and missed visits (e.g., due to illness, changes in travel plans), some children were seen in a wider age window, up to 8 months of age. Infants develop rapidly during the first year, and so for the purpose of this analysis, a narrow, optimal age window was defined as from 1 week before age 6 months to 3 weeks after age 6 months. There were 134 infants who were in this optimal age window (98 high-risk, 36 low-risk), and all results are based on this group. Of note, the wider age group at the first time point will be included in later longitudinal analyses of these data.
Subjects were characterized as having high risk if they had an older sibling with a diagnosis of an ASD that was documented in a clinical diagnostic report and confirmed by the Autism Diagnostic Interview-Revised (18) administered at enrollment. Subjects were enrolled in the low-risk group if they had an older sibling without evidence of ASDs and no family history of a first- or second-degree relative with an ASD. Exclusion criteria for both groups included the following: 1) diagnosis or physical signs strongly suggestive of a genetic condition or syndrome reported to be associated with ASDs (e.g., fragile X syndrome), 2) significant medical or neurological condition affecting growth, development, or cognition (e.g., CNS infection, seizure disorder, congenital heart disease), 3) sensory impairment, such as vision or hearing loss, 4) low birth weight (<2,000 g) or prematurity (<36 weeks gestation), 5) possible perinatal brain injury from exposure to in utero exogenous compounds reported to likely affect brain adversely in at least some individuals (e.g., alcohol, selected prescription medications), 6) non-English-speaking family, 7) contraindication for MRI (e.g., metal implants), 8) adoption, and 9) family history of intellectual disability, psychosis, schizophrenia, or bipolar disorder in a first-degree relative.
The infants were assessed directly at ages 6, 12, and 24 months, and a telephone interview was conducted with the parents at 18 months to assess their infant's development. The direct assessments included brain MRI scans in addition to a battery of behavioral and developmental tests. The assessment battery for the visit at 6 months included the Mullen Scales of Early Learning (19), the Vineland Adaptive Behavior Scales-II (20), the Autism Observation Scale for Infants (21), various questionnaires examining behavior, temperament, and family characteristics, and a medical record review. See Table 1 for a summary of subject characteristics.
Characteristics of 6-Month-Old Infants at Low or High Familial Risk for Autism
| Add to My POL
|Characteristic||Low Risk (N=36)||High Risk (N=98)||Difference|
| Not answered||5||14||4||4|
| Less than college degree||10||28||36||37|
| College degree||8||22||41||42|
| Graduate degree||13||36||17||17|
|Family annual income||0.76|
| Not answered||7||19||10||10|
|Mother's age at infant's birth (years)||32.9||5.7||33.7||4.7||0.37|
|Gestational age at birth (weeks)||39.5||1.2||39.0||1.3||0.47|
|Age at MRI scan (months)||6.3||0.2||6.2||0.3||0.37|
|Mullen Scales of Early Leaning composite score from MRI visit||100.0||10.0||95.2||12.5||0.13|
All the brain MRI scans were completed at each of the clinical sites on a 3-T Siemens Tim Trio scanner with a 12-channel head coil. All scans were completed while the infants were naturally sleeping. A specific structured preparation was completed by the families at home before the scanning; it included conditioning to the scanner sounds on a compact disc played to the infants while sleeping. The imaging protocol was designed to maximize tissue contrast for volumetric analysis across three time points (ages 6, 12, 24 months). The protocol included 1) a localizer scan, 2) a three-dimensional T1 magnetization-prepared rapid acquisition gradient-echo scan: TR=2,400 ms, TE=3.16 ms, 160 sagittal slices, field of view (FOV)=256 mm, voxel size=1 mm3, 3) a three-dimensional T2 fast spin echo scan: TR=3,200 ms, TE=499 ms, 160 sagittal slices, FOV=256 mm, voxel size=1 mm3, and 4) a 25-direction diffusion tensor imaging scan: TR=12,800 ms, TE=102 ms, slice thickness=2 mm isotropic, variable b value=maximum of 1,000 s/mm2, FOV=190 mm.
A number of quality control procedures were employed to assess scanner stability and reliability across sites, time, and procedures. A LEGO phantom (22) was scanned monthly at each location and analyzed to assess image quality and quantitatively address site-specific regional distortions. Two adult subjects (“human phantoms”) were scanned once per year per scanner (twice in year 1). The data for these phantoms were evaluated for scanner stability across sites and time (23). The results indicated excellent stability across sites, with covariates of variation for intracranial volume below 1% and with intraclass correlations for intracranial volume at 0.98 for intersite and 0.99 for intrasite reliability. Finally, all scans were blindly reviewed for image quality by a single rater (D. Louis Collins, McGill University) and again rated by a single reviewer (Rachel Gimpel Smith, University of North Carolina) in a preprocessing stage before image analysis.
All scans were reviewed locally by a pediatric neuroradiologist for radiologic findings that, if present, could be communicated to the participants. In addition, a board-certified pediatric neuroradiologist (R.C.M., Washington University) blindly reviewed all MRI scans across the IBIS network and rated the incidental findings. A third neuroradiologist (D.W.W.S., University of Washington) provided a second blind review for the Washington University site and contributed to a final consensus rating if there were discrepancies between the local site reviews and the network review. The final consensus review was used to evaluate whether there were group differences in the number and/or type of incidental findings.
The following brain volumes were obtained: intracranial, total brain (gray matter plus white matter), total CSF, cerebrum, cerebellum, and lateral ventricles. There is minimal tissue contrast in the 6-month-old brain because of ongoing myelination and maturation. As a result, reliable separation of white matter and gray matter tissue is highly limited. Thus, we chose to initially focus on the total brain volume measure without separating white and gray matter. We also parcellated the brain into large regions (e.g., cerebrum, cerebellum, hemispheres) but omitted a fine-grained lobar or cortical segmentation, which is less reliable at this young age.
The brain volume segmentations were obtained following a rigid transformation to a normative brain space and preprocessing for bias correction and intensity normalization. These steps are part of the expectation-maximization (EM) tissue segmentation method (24) using the AutoSeg tool (25). Tissue probability maps for white matter, gray matter, and CSF were obtained. Intracranial volume was defined to be the sum of white matter, gray matter, and CSF, and an example of the segmentation labeling is shown in Figure 1.
Brain Segmentation at 6 Months of Age
Measures of head circumference were obtained by using a semiautomated image processing tool according to an established protocol (9). Head circumference was measured by a single rater using the HeadCirc tool developed by collaborators at the University of North Carolina. HeadCirc computes head circumference by extracting the appropriate contour from head segmentation of MRI images (using Fourier harmonics to parameterize the contour of the head on MRI). Intra- and interrater reliability for the head circumference measures in the data set from the 6-month-old infants was exceptional, with an intraclass correlation (ICC) of 0.99 for each one. The ITK-SNAP tool (26) was used by a single rater to obtain measurements of the lateral ventricles. ICC values for intra- and interrater reliability for the lateral ventricles were 0.99 (for combined volume and for right and left hemispheres). The AutoSeg, HeadCirc, and ITK-SNAP tools are available at http://www.ia.unc.edu/dev/download and http://www.nitrc.org.
Scans from 12 infants were excluded because they did not pass the initial scan review and/or preprocessing steps. The problems included misalignment (N=2), off-center or distorted image (N=2), poor registration (N=3), and poor image quality or artifact (N=5). There were no group differences in the rate or type of scan failures.
The design of the study was based on findings from our previous comparison of 2-4-year-olds with ASDs and typically developing comparison subjects (10). The percentage difference in the volume of relevant structures (e.g., cerebral cortex, cerebral cortical white matter, and temporal lobe white matter) was 4% to 9%, with effect sizes (Cohen's d) from 0.59 to 0.95, in cross-sectional groups of 45–51 children with ASDs and 15–26 comparison subjects. With our current groups of 36 low-risk and 98 high-risk infants, we expected to have 85%–99% power to achieve effect sizes similar to those in the 2–4-year-old children with ASDs.
Our research goal was to examine whether there were significant group differences in head circumference, brain volumes, and radiologic findings in a group of infants at high genetic risk for autism. Group differences in sex were tested with Fisher's exact test, and separate analysis of variance (ANOVA) models were used to assess group differences in the Mullen early learning composite score, gestational age at birth, age at MRI scan, and maternal age at birth while controlling for site differences. Cross-sectional group differences in brain volumes were analyzed by using a general linear model in SAS (SAS Institute, Cary, N.C.). Covariates included in the model were age at scan, sex (male/female), site, and the Mullen early learning composite standard score. The dependent variables of interest included head circumference, intracranial volume, and volumes of the total tissue (gray plus white matter), CSF, cerebrum (left and right total tissue plus CSF), cerebellum (left and right total tissue plus CSF), and left and right lateral ventricles. To test the influence of sex differences on group effect, we included the additional interaction term (group by sex) in the model.