1. A longitudinal study of brain development in autism
Heather Cody Hazlett, PhD
Neurodevelopmental Disorders Research Center
& UNC-CH Dept of Psychiatry
NA-MIC AHM Salt Lake City, UT Jan 11, 2007

2. Overview Summary of structural imaging studies of autism
Findings from our longitudinal autism study
Challenges & benefits to imaging across development
Future projects & goals for NA-MIC

3. Structural Imaging in Autism

4. MRI Studies of Brain Volume in Autism
Study Brain Finding Subject Age
Piven et al. (1992) mid-sagittal area yrs
Piven et al (1995) total brain volume 14 – 29 yrs
Courchesne et al (2001) cerebral. gray and white 2 – 4 yrs only
Sparks et al (2002) total cerebral yrs
Aylward et al (2002) TBV (HFA) under 12 yrs
Lotspeich et al (2004) cerebral gray (N=52) 7 – 18 yr
Herbert et al (2004) cerebral white 5 – 11 yrs
Hazlett at al (2005) gray matter volume yrs
Palmen et al (2005) TBV, cerebral gray (N=21) 7 – 15 yrs
Limitations: no developmental studies, heterogeneity of samples

5. When compared to typically developing individuals….
increased brain weight in autism
macrocephaly in 20%
increased brain volume on MRI
enlarged tissue volumes (both WM & GM)
age effects present
Train of replicated studies, converging evidence for increased brain size
Recent study shows implications for age effects

6. Longitudinal MRI study of brain development in autism

7. Longitudinal MRI study of brain development in autism
AIMS
To characterize patterns of brain development longitudinally in autism cases versus controls (TYP, DD)
To examine cross-sectional & longitudinal relationships between selected brain regions and behavioral characteristics associated with autism

8. UNC Longitudinal MRI Study of Autism
N % male years (SD) IQ-SS (SD)*
Autism % (0.3) (9.4)
Controls 25
DD % (0.4) (9.4)
TYP % (0.4) (18.7)
* IQ-SS = Mullen composite Standard Score
Hazlett et al Arch Gen Psych 2005

9. UNC Longitudinal MRI Study of Autism
autism controls
mean (SE) mean (SE) % diff p
TBV (13.4) (16.2)
cerebrum (10.5) (12.3)
cerebellum (1.5) (2.2)
Adjusted for Gender and Age

10. UNC Longitudinal MRI Study of Autism
autism controls
mean (SE) mean (SE) % diff p
TBV (13.4) (16.2)
cerebrum (10.5) (12.3)
gray (7.7) (8.8)
white (3.1) (3.7)
cerebellum (1.5) (2.2)

12. UNC Longitudinal MRI Study of Autism
autism typical
mean (SE) mean (SE) % diff p
cerebrum (10.5) (17.4)
gray (7.7) (12.2)
white (3.1) (5.4)
autism dev delayed
mean (SE) mean (SE) % diff p
cerebrum (10.5) (17.2)
gray (7.7) (12.4)
white (3.1) (5.1)

13. Substructures of interest

14. Relationship between Brain Volume and Autistic Features
Social
Communication
Atypical
Behaviors

15. Substructures of interest
Basal ganglia
Caudate
Putamen
Globus pallidus
Amygdala
Hippocampus

16. Caudate Enlargement in Autism
age t p
Study 1
autism
controls
Study 2
autism 15 m =
controls 15 m = 30.3
(Sears, Vest, Bailey, Ransom, Piven 1999)

17. Clinical Correlates of Caudate Volume
ADI Domain Spearman r p
social ns
communication ns
ritualistic/repetitive
(Sears, Vest, Bailey, Ransom, Piven 1999)

18. Clinical Correlates of Caudate Volume
Hollander et al Biological Psychiatry Correlation with total repetitive behavior items on ADI-R
Hollander et al. Biological Psychiatry 2005

19. Descriptives Group N Male M (SD) M (SD) M (SD)
% Years Cognitive* Adaptive**
Group N Male M (SD) M (SD) M (SD)
autism % 2.7 (0.3) (9.3) 60.8 (5.9)
controls % 2.6 (0.5) (28.6) (21.1)
developmental delay 12 67% 2.8 (0.4) (6.7) 65.8 (14.0)
typically developing 21 71% 2.4 (0.5) (16.8) 98.3 (13.4)
* Cognitive estimate from Mullen Composite Standard Score
** Adaptive behavior estimate from Vineland Adaptive Behavior Composite

20. Basal Ganglia Volumes in 2 Year Olds with Autism (adjusted for TBV)
Aut v Total Controls Aut v TYP Aut v DD
diff (SE) p % diff (SE) p % diff (SE) p %
Caudate
.50 (.29) % (.31) % (.43) %
Globus Pallidus
.16 (.29) % (.10) % (.12) %
Putamen
-.16 (.20) % (.22) % (.25) %
Note - all comparisons also adjusted for age and gender

21. Clinical Correlates of Basal Ganglia Volume in 2 year olds with Autism
Caudate Globus Pallidus Putamen
B (SE) p* B (SE) p B (SE) p
ADI Item
Minor Change -.35 (.230) (.071) (.135) .001
Rituals
Body Mvt .413 (.150) (.049)
* one-sided t-test

22. MRI Studies of Amygdala Volume in Autism
Sparks (2002) 45 ASD inc vs. TYP and DD controls
(3-4 yr olds)
Schumann (2004) 61 ASD increased in 7-12 year olds,
not increased year olds
Schumann et al Journal of Neuroscience Cross sectional study. No difference in total cerebral volume between Aut v Typ groups. Aut children had larger left and right AMYG v. typicals but there were no differences in AMYG for older children Note that AMYG in TYP children increases in volume between 7.5 and 18.5, while children with Aut have initial enlargement but don’t show this rapid increase.

23. Amygdala/Hippocampus Volume in 2 Year Olds with Autism
(adjusted for TBV)
Aut v Total Controls Aut v TYP Aut v DD
diff (SE) p % diff (SE) p % diff (SE) p %
amygdala
.35 (.12) % (.11) < % (.17) %
hippocampus
.03 (.11) % (.14) % (.15) %
*Note – all comparisons also adjusted for age and gender

24. FXS-autism vs autism-nonFXS
FXS (N=35); Controls (N=38); FXS + autism (N=12); Autism - nonFXS (44)

25. Imaging Development

26. Challenges to Developmental Studies
Difficult for very young children and/or lower functioning children to remain still
May need to remain motionless for long periods of time
Sleep studies vary in success rates
Subjects may require training and practice – this adds to expense

27. To total cerebral white matter Longitudinal Studies:
Brain Development During Childhood and Adolescence
total cerebral
frontal gray parietal gray
Longitudinal Methods
time 1  time 2
12 yrs
12 yrs
more sensitive for detecting growth patterns, even in the presence of large inter-individual variation and non-linear growth
Peak 12 y
temporal gray occipital gray
Giedd et al Nature Neuroscience scans (normals) from 89 males, 56 females, ages 4-22 yrs. Males purple, females red (displayed with confidence intervals)
16 yrs
20 yrs
Age in years 4
Giedd et al., Nature
Neuroscience, 1999

28. Gray matter maturation
N. Gogtay, Giedd et al PNAS Depicts gray matter maturation over the cortical surface, from age 5-20 yrs. Very small sample of 13 (7 male, 6 female) normal subjects.
Gogtay, Giedd et al PNAS N = 13 (7 male, 6 female) typical subjects

29. Time Course of Critical Events in the Determination of Human Brain Morphometry
Neurodevelopmental processes, cortical synapse density, and their relationship to gray and white matter volumes on MRI. Giedd et al. 1999, Sowell et al

30. Neonatal Brain MRI T2 T1 gray matter non-myelinated white matter
early myelinated white matter T2 T1

31. Corpus Callosum Neonate (2 wks) Infant (1 year) Adult
Corpus callosum: FA along Commissural bundles

32. Infancy to Childhood Hermove et al., NeuroImage 2005.
Hermoye et al… (Mori) NeuroImage Subjects were 7 healthy and 23 pediatric patients (17 boys, 13 girls)
Hermove et al., NeuroImage 2005.

33. Data

34. Data Structural MRI Diffusion Tensor
Behavioral, cognitive, developmental
Processed longitudinal data

35. Data
Structural MRI
TI: coronal 3D SPGR IRprep, 0.78 x 0.78 x 1.5 mm, 124 slices, 5 TE/12 TR, 20 FOV, 1 NEX, 256x192
PD/T2: coronal FSE, 0.78 x 0.78 x 3.0 mm, 128 slices, 20 FOV, 17 TE/7200 TR, 1 NEX, 256x160
DTI
axial oblique 2D spin echo EPI, 0.93 x 0.97 x 3.8 mm, 30 slices, 24 FOV, 12 dir
T1: 3D SPGR IR Prep, coronal, mm x mm x 1.5 mm, 124 slices, field of view = 20 cm x 20 cm scan time = 7:10, NEX = 1 flip angle = 20, TE = 5, TR = 12, TI = 300, acquisition matrix = 256 x 192 PD/T2: 2D Fast Spin Echo (FSE), coronal, mm x mm x 3.0 mm, interleave acquisition, slice number to cover brain (usually 128 slices), field of view = 20 cm x 20 cm scan time = 9:36, NEX = 1 TE = 17, 75, TR = 7200, ETL = 8, acquisition matrix = 256 x 160 DTI: 2D Spin Echo-EPI, axial oblique (to ACPC), mm x mm x 3.8 mm, 0.4 mm gap, 30 slices, field of view = 24 cm x 24 cm, 4 acquisitions, baseline plus 12 directions (6 directions and their inverses) scan time = 2:38 per acquisition (10:32 total) TE = min, TR = 12200, acquisition matrix = 128 x 128

36. Data Processed datasets* Time1 (2 yr old) Time2 (4 yr old)
EMS/lobes CN AMYG EMS/lobes CN AMYG
Autism (+2 CS) DD
Typical FX
Also have segmented data for:
Put/GP, Hipp, CC area, Ventricles, Ant Cing
*As of Nov06

37. Image Processing

38. Tissue segmentation –2 yr old
EMS hard segmentations
EMS segmentations overlaid on MRI
1)      The registered T1, T2, and PD images were used as input to the EMS program. EMS, or Expectation Maximization Segmentation, is a program for automated tissue classification and brain stripping. (we probably need an EMS reference) This program makes use of a template image, which is a fuzzy/averaged image. This template image is used in the segmentation process to provide spatial information of tissues: It is a probability map for where the different tissues are most and least likely to be located.
2)      EMS gives us two different kinds of output images: One is a soft segmentation; one is a hard segmentation. The soft segmentation is like a probability image for a particular tissue for a particular subject. We get separate soft segmentation images for GM, WM, and CSF (as well as background images). The hard segmentation (our output combines GM, WM and CSF hard segmentations into a single label image) is a discrete tissue classification, in which there are hard boundaries between tissue types.
3)      We use the hard, or discrete, segmentation to obtain ICV (intracranial volume). This is used as co-variate for the CN volumes (as well as a comparison of overall brain size among groups). Our ICV value includes GM, WM, and CSF. The hard segmentation goes through a “brain stripping” process to isolate the brain and surface CSF from surrounding tissues. This is done in part by the use of the template image and in part by an integrated post-processing step (occurs at the same time the GM, WM, and CSF hard segmentations are combined into a single image) in which small, unconnected pieces are excluded.

39. Automatic parcellation by template warping
Manually-derived parcellation “warped” to new subjects

40. Challenges to Image Processing

41. Challenges to Image Processing
Registration of images to a common atlas
Inhomogeneities – bias correction
Tissue contrast – myelination
Brain shape changes across development

42. Future Directions

43. Future Directions Examination of longitudinal data
e.g., 2-4 years old, follow-ups at 6-8
Development & application of novel image processing methods
e.g., shape, cortical thickness

44. Change from 2 to 4 years
These frames show the evolution from 2 year old to 4 year old using
high dimensional fluid warping (Joshi)

45. Surface growth maps age 2 4
Blue = growth, red = atrophy, green = static
age

46. NA-MIC Collaboration Goals/Projects for NAMIC collaborators:
Pipelines for growth-rate analysis
Longitudinal analysis of cortical thickness, cortical folding patterns, etc.
Automating DTI processing, creating more regionally defined DTI analysis (?)
Development of new segmentation protocols (e.g., dorsolateral prefrontal cortex)
Quantify shape changes over time to allow for analysis with behavioral data

47. NA-MIC Collaboration Our site can offer NAMIC collaborators:
Pediatric dataset of sMRI & DTI
Longitudinal data
Segmented datasets (e.g., substructures, ROIs) to be used as validation tools

48. Contributors Martin Styner, PhD Allison Ross, MD Guido Gerig, PhD
James MacFall, PhD
Alan Song, PhD
Valerie Jewells, MD
James Provenzale, MD
Greg McCarthy, Ph.D.
John Gilmore, MD
Allen Reiss, MD
UNC Fragile X Center
NDRC Research Registry
Funded by the National Institutes of Health
Joe Piven, MD
Guido Gerig, PhD
Sarang Joshi, PhD
Michele Poe, PhD
Chad Chappell, MA
Judy Morrow, PhD
Nancy Garrett, BS, OTA
Robin Morris, BA
Rachel Smith, BA
Mike Graves, MChE
Sarah Peterson, BA
Matthieu Jomier, MS
Carissa Cascio, PhD
Matt Mosconi, PhD
Many thanks to the families that have generously participated !