fMRI of Normal Language Development in Children

1
fMRI of Normal Language Development in Children
Scott K. Holland, Ph.D. Professor of Radiology,
Pediatric Physics, Neuroscience and Biomedical
Engineering President, 3Tesla Consulting, LLC
Imaging Research Center Cincinnati Childrens
Hospital
2
A Functional Magnetic Resonance Imaging (fMRI)
Study of Language Development

• Began with studies of language lateralization in
  pediatric epilepsy patients in 1997.
• Discovered wide variations in language
  representations.
• Due to age and development?
• Due to pathology?
• Due to fMRI?

3
Study Population

• Study inclusion criteria
• Age ??18 years
• Seizures possible temporal lobe onset
• No sedation required
• 20 pediatric epilepsy patients completed fMRI
  language exams
• Average age 13.3 yrs
• Range 8-18 yrs

4
Methods for Language Lateralization in Pediatric
Epilepsy circa 1998

• Verb generation task
• Block periodic design
• 1.5 Tesla, GE Signa 5.7, EPIBOLD
• Hemispheric LI from thresholded correlation maps
• LI Right Pixels - Left Pixels
• Right Pixels Left Pixels

AES '99 Pediatrics Symposium
5
EPIBOLD - fMRI of Language at 1.5 T
Sagittal Slice Locations in Temporal Lobes for
Language Lateralization
608607/1325 - KD
6
Composite activation map for pediatric epilepsy
patients
Left Hemisphere
Video presentation (yellow, N17) Audio
presentation (red, N15). Orange pixels Video
Audio
Right Hemisphere
Verb generation paradigm at 1.5 Tesla. Hot pixels
correspond to a cross-correlation value between
0.2 and 0.4 with cluster size 3.
7
Results Data Table
Cross-Correlation Map (R 0.5, Clusters 3)
girls
boys
Average Age 13.3 years
Red gt audio L.I. and video L.I. disagree Blue gt
audio L.I. and video L.I. agree
8
Lateralization with Age in Epilepsy Patients
Holland, S.K., Strawsburg, R.H. and Weber, A.M.,
"Functional Magnetic Resonance Imaging of
Children with Epilepsy". American Epilepsy
Pediatric Symposium Brain Imaging Techniques in
Children with Epilepsy, Orlando, 1999.
9
What we learned -Importance of Normative fMRI
Data

• fMRI assessment of language dysfunction in
  children must account for normal developmental
  differences.
• Few fMRI data relating to normal language
  development
• novelty of applying fMRI in children.
• difficulty obtaining cooperation of normal young
  subjects.
• must be done without sedation.
• Methodology is important for consistent result
• High quality acquisition
• Robust post-processing
• Large subject population

10
Earlier fMRI Studies of Language

• Focused mainly on language disorders.
• Epilepsy
• Dyslexia
• Tumors
• Stroke
• Specific language impairments
• Normal variations had not been mapped.
• As function of age
• As function of development and language ability
• As function of language domain

11
Developmental Aspects of Language

• Language is not a static process.
• How do neural substrates of language change
  throughout development?
• Does hemispheric lateralization of language
  function change with age?
• How is the developmental trajectory affected by
  injury?

12
The American Colleges and the American Public,
1870,
13
The CNS (Cincinnati Navigational System)
N
Cincinnati
14
Language Components Examined

• Phonologic
• Orthographic
• Semantic v
• Syntactic v

15
Rationale for Language Paradigms
Syntax (Sentential)
Semantic (Lexical)
Early Developing
1) Syntactic Prosody
3) Picture Matching
Later Developing
2) Story Processing
4) Verb Generation
16
Early Developing Syntactic Task 1) Syntactic
Prosody

• Your sentence is So the frog, who missed his
  flower, went inside.
• _at_ !_at_( )!?" lt/ !_at_ ?lt-
  )gt\lt,?_at_..
• Th fro a und he R f lo wert.
• ! lt) )?? \ !_at_( )!?"!
• Th fro a und he R f lo wert.
• Th fro a und he R f lo wert.
• !_at_ ?lt- (-\ _at_gt ?! ! )gt\lt,?_at_.
• (-\ _at_gt ?! ! ) (-\ _at_gt ?! ! \
  ."lt.
• Th fro a und he R f lo wert.
• ? _at_) lt! ? lt) )?? \ - ?gt!_at_.
• ? _at_) lt! ? (-\ _at_gt lt) )?? \.

Control Tone identification
17
Later Developing Syntactic Task 2) Story
Processing

• A very naughty monkey lived in the city zoo.
• Once, he learned how to open his own cage.
• He teased the tiger and stole his food bowl.
• The monkey saw a lady with a shiny necklace.
• If he was sneaky, he could steal it too!
• He grabbed the necklace and ran up a tree.
• The lady screamed, Who stole my pretty gold
  necklace?
• The zookeeper saw that it was the naughty
  monkey.
• Maybe we can trade for it, the zookeeper
  said.
• Finally, the monkey dropped the necklace for a
  banana.

Control Random Tones
18
Early Developing Semantic Task 3) Picture
Matching
Fish
19
Early Developing Semantic Task 3) Picture
Matching Control
Tone
20
Later Developing Semantic Task 4) Verb
Generation Task

• noun verb
• (person, place, or thing) (action word)
• ball throw
• hit
• kick

Control Bilateral finger tapping
Holland, S.K., Plante, E., Weber, A.M.,
Strawsburg, R.H., Schmithorst, V.J. and Ball,
W.S., Jr., "Normal Brain Activation Patterns in
Children Performing a Verb Generation Task".
Neuroimage, 14837-43, 2001.
21
Study Population

• Group I - 280 normal children
• 10 Girls and 10 boys 20 children
• Each year of age 5 - 18
• 190 recruited to date
• Group II - 10 longitudinal subgroup
• 7 year olds
• Return each year for 5 years
• Group III - 20 left MCA stroke
• Perinatal left MCA stroke
• Abnormal language patterns

22
Inclusion Criteria

• Between 5 and 18 years of age
• 5th to 95th percentile weight and height for age
• IQ and OWLS scores w/i 1s
• Normal physical neurological examination
• Normal head circumference
• At least C-minus average in school
• Negative history of disorder
• Neurological
• Psychological
• Neuropsychiatric ADHD, TS, LD, etc.
• Informed consent/assent

23
Exclusion Criteria

• Standard MRI exclusion criteria
• Metallic implants that interfere with MRI
• Special education placement
• Full Scale IQ lt s on WISC-III or WAIS-III
• Previous history of head trauma
• Gestational age lt 36 weeks
• Birth weight lt 25th percentile
• Pregnancy
• Orthodontic braces

24
Susceptibility Effect of Orthodontic Braces
Devastates fMRI Signal at 1.5 T
838333/01530 - DG
15 y.o. male w/ braces history of seizures -
bilateral finger tapping
25
MRI Scans

• 3.0 Tesla, Bruker BioSpec
• BOLD fMRI with EPI GRE
• Block Periodic Task Designs
• Diffusion Tensor Imaging
• 3D T1 Anatomical images
• 1.5 x 1.0 x 1.5 mm resolution

26
fMRI Scans

• 3.0 Tesla, Bruker MedSpec
• 100 EPI gradient echo, single shot images
• 2000/38 msec TR/TE
• 25.6 cm/64x64 FOV/Matrix
• 5/0 mm Slice thickness/Gap
• 24-32 transverse slices
• Total acquisition time 5.5 minutes
• Reconstruction in within 10 minutes

27
fMRI Failure Rates by Age Sex
Byars, A.W., Holland, S.K., Strawsburg, R.H.,
Bommer, W., Dunn, R.S., Schmithorst, V.J. and
Plante, E., "Practical Aspects of Conducting
Large-Scale Functional Magnetic Resonance Imaging
Studies in Children". J Child Neurol, 17885-90,
2002.
28
Recruitment as of 2/19/04 - BOYS
29
Recruitment as of 2/19/04 - GIRLS
30
Success with fMRI kids means understanding their
tastes
31
5 y.o. Girl - Verb Generation fMRI Scan
3T, Verb Generation, Rgt0.45, cluster size7
32
Socioeconomic Status by Census Tract
SES Distribution, N351 as of 11/30/03
33
Full Scale IQ (Mean 112 11, N329, 11/20/03)
11/20/03
34
Normalization of Pediatric Brain Image Data
Which Standard?

• Adult Standard - Talairach and Tournoux
• Co-Planar Stereotaxic Atlas of the Human Brain
  3-Dimensional Proportional System An Approach to
  Cerebral Imaging, Thieme, 1988
• AC-PC reference points
• standard x,y,z dimensions
• Does it work in children?
• ICBM?

35
Landmarks of Proportional Grid System Talairach
and Tournoux
PC
AC
Jean Talairach and Pierre Tournoux, Co-Planar
Stereotaxic Atlas of the Human Brain
3-Dimensional Proportional System An Approach
to Cerebral Imaging, Thieme, 1988
36
Normalization of Child Brain DataPediatric vs.
Adult Templates
Wilke, Schmithorst, Holland, Human Brain Map,
Vol 17(1) 2002
37
Wilke et al Figure 37
Normalization of Pediatric Brain Image Data
full normalization
Determination of the Deformation field
Point-to-Point tracking
Non-linear transformation (SPM-template)
non-linear only
Affine transformation
full normalization
Non-linear transformation (CHMCC-template)
Determination of the Deformation field
Point-to-Point tracking
Parameters of linear transformation
non-linear only
38
Affine Transformation Scaling Factors for
Pediatric --gt Talairach
N 147, Age range 5-18 years
39
Wilke et al Figure 39
a.
b.
40
Conclusion - Normalization

• Talairach reference frame appears adequate for
  use in children for spatial resolution maps of 3
  mm or more.
• ( M Wilke, VJ Schmithorst,SK Holland, Human
  Brain Map, 17(1) 2002)
• Below 5 mm resolution, misregistration may become
  an issue. (ED Burgund, HC Kang, JE Kelly, RL
  Buckner, AZ Snyder, SE Petersen, BL Schlaggar.
  NeuroImage 17, 184-200 (2002)
• Talairach is untested in children under 5 y.o.
  and we expect increasing errors with infants and
  toddlers.
• (SK Holland, D Choo, VJ Schmithorst. 9th
  Symposium on Cochlear Implants in Children,
  Baltimore, (2003)
• Pediatric templates available in MNI frame.
• (Wilke, M., Schmithorst, V.J. and Holland, S.K.,
  "Normative Pediatric Brain Data for Spatial
  Normalization and Segmentation Differs from
  Standard Adult Data". Magn Reson Med, 50749-57,
  2003) - http//www.irc.cchmc.org

41
Composite Activation Maps chloral hydrate
sedated, normal hearing children (ave. age
11.6 mo.)

• Tone Task (2/5) Story Task (2/5)
• Group Composite Activation Map (n 5)
• Z-score, corrected plt0.05

42
Data Representations

• Composite maps
• Correlation parameter maps
• Evolution Videos
• Lateralization growth curves

43
Syntactic Prosody N259 Group Composite
Activation Map Z-score, corrected plt0.001
4/23/03
44
Story Processing N254 Group Composite
Activation Map Z-score, corrected plt0.001
4/23/03
45
Picture Matching N262 Group Composite
Activation Map Z-score, corrected plt0.001
4/23/03
46
Verb Generation N257 Group Composite Activation
Map Z-score, corrected plt0.001
4/23/03
47
Correlation Between Activation Level and Age
Syntactic Prosody Task, N259
4/23/03
48
Correlation Between Activation Level and Age
Story Processing Task, N254
4/23/03
49
Correlation Between Activation Level and Age
Picture Matching Task, N262
4/23/03
50
Correlation Between Activation Level and Age Verb
Generation Task, N257
4/23/03
51
Lateralization Index Calculation

• Average Z-score within symmetric ROIs.
• Generalized linear model (GLM)
• Group composite average map
• Low Z -score threshold ROIs
• LI SZleft - SZright / SZleft SZright
• Immune to thresholding effects

52
Examples of Regions of Interest for Verb
Generation
QuickTime and a
TIFF (LZW) decompressor
are needed to see this picture.
Wernickes ROI
Brocas ROI
53
Lateralization Index vs. AgeStory Processing
Task (N225)
Lateralization Index
Age in months
4/23/03
54
Lateralization Index vs. AgeVerb Generation Task
(N228)
Lateralization Index
Age in months
4/23/03
55
Lateralization Index vs. AgeFinger Tapping
(N228)
Lateralization Index
Age in months
4/23/03
56
Discussion

• Task design is consistent with model of early and
  later developing language skills.
• Syntactic language functions are bilaterally
  distributed across age span tested.
• Semantic language functions are left lateralized
  across age span tested.
• Later developing semantic skills (lexicon) show
  greater left lateralization increases.

57
Discussion

• Focus of semantic language function appears to
  increase in left frontal lobe (Brocas Area) with
  age.
• Focus of syntactic language function may emerge
  in right temporal lobe auditory areas with
  increasing age.

58
Questions

• Does left lateralization continue to increase
  throughout adulthood?
• How does our pediatric data merge with exiting
  literature on decreasing left lateralization with
  age in adults?

59
Verb Generation LI vs. Age 5-65 years
0-5 y.o.
60
Longitudinal Population - 4 visits
Pixels where slope of Z-score vs. Age differs
from 0 with p lt 0.001 5th visits will be
completed in 2004.
Story Task
Verb Task
61
Volumetric Correlations using Voxel Based
Morphometry

• Correlations between GM and WM volume and age.
• Wilke, M. and Holland, S.K., "Variability of
  Gray and White Matter During Normal Development
  A Voxel-Based MRI Analysis". Neuroreport,
  141887-90, 2003.
• Correlations between GM and WM volume and IQ.
• Wilke, M., Sohn, J.H., Byars, A.W. and Holland,
  S.K., "Bright Spots Correlations of Gray Matter
  Volume with IQ in a Normal Pediatric Population".
  Neuroimage, 20202-15, 2003.

62
Age related changes in variability of brain
volumes in GM (left) and WM (right) based on VBM
Blue-Green higher variability in older
children, Red-Yellow higher variability in
younger children
Wilke, M. and Holland, S.K., Neuroreport,
141887-90, 2003.
63
Gray Matter Volume vs. IQ
After correction for age, gender and handedness,
GM volume is plotted against IQ in cingulate
gyrus (blue dots, red line) and entire cerebrum
(gray diamond, gray line) for n200 pediatric
subjects 5-18 years of age.
64
Bright SpotsCorrelation of GM Volume with IQ
Connectivity Analysis in SPM99 - VBM analysis
showing regions where GM volume correlation with
IQ follows the same pattern as regions that
significantly correlate with IQ (n200). Red
positive correction and Green negative
correlation with IQ.
Wilke, M., Sohn, J.H., Byars, A.W. and Holland,
S.K., Neuroimage, 20202-15, 2003.
65
Diffusion Tensor Imaging

• Correlations between diffusion parameters
  (fractional anisotropy and trace) and age.
• Schmithorst, V.J., Wilke, M., Dardzinski, B.J.
  and Holland, S.K., "Correlation of White Matter
  Diffusivity and Anisotropy Changes with Age
  During Childhood A Cross-Sectional Diffusion
  Tensor Imaging Study". Radiology, 222212-8,
  2002.
• Correlations between diffusion parameters and IQ.
• Schmithorst, V.J., Wilke, M., Dardzinski, B.J.
  and Holland, S.K., While Matter Structure
  Correlates with IQ in Normal Children A
  Diffusion Tensor Imaging Study." NeuroReport,
  Submitted2004.

66
Trace ADC changes with age reflects White Matter
Changes

• DTI in 33 normal, healthy children

• Trace(D) shows a strong, negative correlation
  with age

• could reflect an increase in tissue density
  via increased association fibers

67
FA-changes with age reflects White Matter Changes

• DTI in 33 normal, healthy children

• FA shows a strong, positive correlation with
  age

Longitudinal fasciculus
Arcuate fasciculus

• could reflect an increase in tissue organizati
  on in primarily projective axons

Internal capsule
Cortico-spinal tract
68
DTI Correlations with IQ
Regions of statistically significant negative
correlations of FA (left) and positive
correlations of TADC (right) with Wechsler
full-scale IQ scores, overlaid on the averaged
whole-brain anatomical dataset from n47
children. Colored voxels have p lt 0.03
(corrected).
69
DTI Correlations with IQ
Regions of statistically significant positive
correlations of FA (left) and negative
correlations of TADC (right) with subject age,
overlaid on the averaged whole-brain anatomical
dataset. Colored voxels have p lt 0.03
(corrected). N47
70
Neuroplasticity and Reorganization Following
Brain InjuryHow does language distribute in the
brain following perinatal deletion of typical
cortical areas?

• Group III - 20 left MCA stroke
• Perinatal left MCA stroke
• Abnormal language patterns
• Group IV - 30 left MCA Stroke
• Infarct age 5-8
• Imaging at 2-7 years post-injury

71
Cortical distribution of the middle cerebral
artery (MCA) encompasses most language areas of
interest.
Heimer, L., The Human Brain and Spinal Cord
Functional Neuroanatomy and Dissection Guide.
1983, New York Springer-Verlag.
72
7 year old boy with perinatal left MCA infarction
7 y.o. Male Pt. Normal 7 y.o. N17
Story Task
Verb Task
73
12 year old girl with perinatal left MCA
infarction
12 y.o. Female Pt. Normal 12 y.o. N11
Story Task
Verb Task
74
Conclusion - Perinatal Stroke

• Perinatal left MCA infarction produces mixed
  contralateral reorganization of language.
• Language abilities are largely restored 6-8 years
  after perinatal stroke.
• Outcome is less favorable for later injuries.
• Can therapies augment contralateral
  reorganization?

75
What about less severe injuries?

• Effects of Perinatal lead exposure
• fMRI study of neural substrates of
• Language
• Attention
• Working memory
• Pb exposure groups
• High gt 20 mg/dl
• Low lt 10 mg/dl

76
Verb Generation Task Pb Exposure
Composite map of n16, normal 18 y.o. subjects
Composite map of n16, 22 y.o. Lead exposed
subjects
77
Verb Generation Task Pb Exposure
GLM random effects analysis shows areas where
normal controls are significantly more active
during verb generation than lead exposed subjects
(n16) and (n16), p lt 0.001
78
Conclusions - Lead Exposure

• Subtle differences are apparent in lead exposed
  subjects when compared to the age matched,
  normative reference data.
• Differences are less severe than in children with
  left MCA stroke.
• Focal differences may provide clues to mechanism
  of brain injury in Pb exposure.

79
Handedness Differences

• N 15 LH subjects
• 73 (n11) left lateralized
• 27 (n4) bilateral
• N 15 RH controls
• 93 (n14) left lateralized.
• Blue pixels represent regions where LH gt RH, but
  none of these pixels reached significance.

80
Conclusions - Handedness

• Effects of handedness on hemispheric
  lateralization of language are subtle.
• Large subject cohorts will be needed to determine
  significance of regional differences.
• Similar effects observations apply to sex
  differences in lateralization.

81
Future Directions

• Improved auditory language stimulation paradigms
  should be developed to minimize confounding
  auditory stimulation from the scanner noise.
• HUSH Hemodynamics Unrelated to Scanner
  Hardware.
• Schmithorst, V.J. and Holland, S.K.,
  "Event-Related fMRI Technique for Auditory
  Processing with Hemodynamics Unrelated to
  Acoustic Gradient Noise". Magn Reson Med,
  51399-402, 2004.
• More sophisticated data analysis methods (e.g.
  ICA) may yield new insights about language
  development from this data.
• Schmithorst, V.J. and Holland, S.K., "Comparison
  of Three Methods for Generating Group Statistical
  Inferences from Independent Component Analysis of
  Functional Magnetic Resonance Imaging Data". J
  Magn Reson Imaging, 19365-8, 2004.

82
HUSH - fMRIA silent event related acquisition
Gradients/Acq.
Acquire
Acquire
Silent
Silent
Silent
Silent
Acquire
Acquire
Auditory Stim.
None
None
M.S.
Silence
M.S.
Silence
Tones
None
6
6
6
5
5
5
2
Interval Time (sec)
Running Time (sec)
0
2
7
13
18
24
29
35
Diagram of 3-phase HUSH-fMRI acquisition sequence
with auditory stimulation during silent-gradient
intervals.
83
HUSH-fMRI in a 1 y.o. male infant
Story Task
Tone Task
1 y.o. infant with Demorsier's Syndrome, Sedated
with 75 mg/Kg chloral hydrate and 3.5 mg/Kg
Nembutal.
84
Independent Component Analysis of Language fMRI
Data
Story Processing N-187
Verb Generation N188
Most highly task correlated component from ICA
85
Independent Component Analysis of Language fMRI
Data
Story Processing
Verb Generation
Task correlated components 2 and 3 from ICA
86
What do the Independent Components represent?
87
Future Directions

• Improved ranking methods for ICA
• Age dependence of the BOLD effect needs further
  investigation to unravel effects of cognitive
  ability, effort, attention, physiological
  changes, etc.
• Age dependence of motion artifact and impact on
  fMRI data needs further study.

88
Age Dependence of BOLD Effect
Talairach composite correlation map for verb
generation and finger tapping tasks showing
pixels that have a percent signal change in the
BOLD effect that correlates with age of the
subjects. (p lt 0.001, red verb generation,
blue finger tapping.
89
Age Dependence of BOLD Effect
90
Age Dependence of BOLD Effect
91
Acknowledgements(funding)

• This work was funded by grants from
• Cincinnati Childrens Hospital Trustees
• NICHD - RO1-HD38578
• NIDCD - R21-DC41018

92
Acknowledgements(Collaborators)

• Anna Weber Byars, Ph.D. - CCHMC
• Richard Strawsburg, MD - CCMC
• Mark Schapiro, MD - CCHMC
• Vincent Schmithorst, Ph.D. - CCHMC
• Antonius DeGrauw, MD, Ph.D. - CCHMC
• Jerzy Szaflarski, MD, Ph.D. - University of
  Cincinnati
• Marko Wilke, MD - University of Tuebingen
• Elena Plante, Ph.D. - University of Arizona

93
Acknowledgements(Students)

• Lori Arlinghaus - Wash. U/ Vanderbilt U.
• Jennifer Frey - Vanderbilt University
• Jennifer Ret - Miami University
• Daniel Shrey - Northwestern U./ U. Cinci.

94
Acknowledgements(Helpful Discussions)

• William Gaillard, MD - Childrens National
  Medical Center George Washington Univ. School
  of Medicine
• Ken Pugh, Ph.D. - Yale University Haskins Labs
• Guenivere Eden, Ph.D. - Georgetown University
• Peter Chiu, Ph.D. - University of Cincinnati
• Peter Fox, MD - Univ. Texas, San Antonio HSC
• Keith Thulborn - University of Illinois, Chicago
• Martin Staudt, MD. - University of Tuebingen