Creditcontributorsgrants

1
Studying the Developmental Neurobiology of
Reading Using fMRI
Bradley L. Schlaggar MD PhD John Merck Fund
Summer Institute, July 23, 2003
2
B.L.S. is a Scholar of the Child Health Research
Center of Excellence in Developmental Biology at
Washington University School of Medicine
(HD33688).
Acknowledgements
Steven E Petersen PhD Tim Brown BS Kristina
Visscher BS Kristin Wenger BS Erica Palmer
MS Christine Kang MD Darcy Burgund PhD Jim Kelly
BS Randy Buckner PhD Avi Snyder MD PhD David Van
Essen PhD
Fran Miezin MS Mark McAvoy PhD Heather Lugar
BS Becky Coalson BS Tara Spevack PhD
NIH NINDS NICHDMcDonnell Center for Higher
Brain Function Charles A Dana
Foundation Burroughs-Wellcome Fund John Merck
Scholars Fund
3
Cytoarchitectonic Map of Human Cerebral Cortex
From Brodmann,1909
4
How and when does the mature organization
emerge?What are the rules?
5
Gross Development of Human Cerebrum
From Cowan,1979
6
Why study typical and atypical brain development?

• Informs normal function
• Developmental disabilities and consequences of
  brain injury
• Generate rational interventions and assay their
  effects

7
A developmental context

• our real teacher has been and still is the
  embryo ---who is, incidentally, the only teacher
  who is always right
• Viktor Hamburger (1900-2001)

8
Importance of Reading

• Straightforward predictor of success in school
  and later in life.
• Understanding how skilled reading is carried out
  and acquired
• critical to improving strategies for reading
  education and
• for identifying reading disabled for early and
  effective remediation

9
Reading

• Taught
• Not evolutionarily driven
• Disabled reading entirely consistent with normal
  intelligence (dyslexia).

10
In general

• Reading is sub-served by a left hemisphere
  network of cortical regions for mapping visual
  (orthographic) information onto
• auditory (phonological) and conceptual
  (semantic) representations.

• frontal
• articulation/word analysis
• parieto-temporal
• orthographic-phonological word analysis,
• occipitotemporal
• word form

11
In general

• Reading is sub-served by a left hemisphere
  network of cortical regions for mapping visual
  (orthographic) information onto
• auditory (phonological) and conceptual
  (semantic) representations.

• Transition to expertise
• Early parieto-temporal word analysis dorsal
  system
• Skilled occipito-temporal word form ventral
  system

12
Approaches/Tools

• Cognitive Psychology
• Behavioral Neurology
• Cognitive Neuroscience/Functional Neuroimaging

13
Initial State Logographic Readingfrom Ramus
Object
Text
Speech
14
Intermediate State Alphabetic Readingfrom Ramus
Object
Text
Sub-lexical Alphabetic Representation
Phonological awareness
Speech
15
Final State Orthographic Readingfrom Ramus
Object
Text
Speech
16
Approaches/Tools

• Cognitive Psychology
• Behavioral Neurology
• Lesion/behavior
• Alexia without agraphia (dom. medial occipital
  and inferior fibers of splenium of corpus
  callosum)
• Alexia with agraphia (dominant angular gyrus)
  Gerstman
• Cognitive Neuroscience/Functional Neuroimaging

17
Approaches/Tools

• Cognitive Psychology
• Behavioral Neurology
• Cognitive Neuroscience/Functional Neuroimaging

18
Functional MRI

• Safe
• FDA approved sequences
• Non-ionizing
• PET, SPECT
• Non-invasive

8 year old volunteer in mock scanner
19
The basis of fMRI is the BOLD Effect (blood
oxygen level dependent)
BOLD Hemodynamic Response Function

• deoxyHg paramagnetic oxyHg not
• Neural activity-gtluxury perfusion
• deoxyHg/ oxyHg decreases
• Given volume less magnetic
• T2 relaxation in given volume slows
• T2 change is the BOLD effect

TIME (SEC)
0
4
14
32
From Josephs, Turner, Friston 1997
20
Conceptualization of the developing neocortex

• Conceptual framework/pendulum
• Nature neocortex is entirely hard-wired at
  birth.
• Nurture neocortex is entirely equipotent.
• Softer versions of these models emerged in the
  1980s
• Rakic protomap (i.e. not a fatemap)
• OLeary protocortex (i.e. not a tabula rasa)
• Analogy with the cognitive development literature
• Chomsky, Fodor, Pinker,(Kanwisher?) nativist,
  modularist
• Elman et al, Quartz Sejnowski, Johnson,
  Karmiloff-Smith connectionist, selectionist,
  neuro-constructivist

21
Potential Developmental Scenarios

• Activation of a nascent adult organization
• nativist
• Intially diffuse organization becomes
  specialized
• selectionist
• Acquisition and skilled performance are
  sub-served by different neural mechanisms
• Neuro-constructivist/scaffolding

22
Regressive (and Progressive) Events in Building a
Brain
Cowan et al Science 1984

• Progressive
• Cell proliferation
• Migration to definitive locations
• Selective aggregation
• Establishment of phenotypic diversity
• Establishment of complex connections
• Neuropil expansion
• Myelination

• Regressive
• Restriction of phenotypic potential
• Cell death
• Pruning of synapses
• Elimination of exuberant connections/processes
• Cowanian model of brain development
• Deprivation of trophic factors

23
Potential Developmental Scenarios

• Activation of a nascent adult organization
• nativist
• Intially diffuse organization becomes
  specialized
• selectionist
• Acquisition and skilled performance are subserved
  by different neural mechanisms
• Neuro-constructivist/scaffolding

24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
Where to begin to study reading development?

• Even automatic/expert reading is very complex
• Requires the coordination of multiple visual,
  oculomotor, and linguistic mechanisms.
• Start with reading words aloud
• Orthography to phonology
• Easy to manipulate experimentally

33
Functional neuroimaging and single word reading

• A wide variety of lexical tasks examined
  requiring not only word reading, but also
  performance of complex operations on single
  words.
• But, relatively little work has been specifically
  dedicated to the functional neuroanatomy of
  single word reading.

34
Neuroimaging studies of skilled word reading

• Variables manipulated
• frequency, regularity, lexicality, letter case,
  word length, stimulus degradation, rate, duration
• Context
• lexical decision, verb past tense generation,
  object naming, simple reading.
• Control tasks (for baseline comparison)
• resting with eyes closed, visual fixation,
  passive viewing of words, silent reading of words
  , uttering a pre-determined word in response to
  consonant strings, false fonts.
• Responses
• vocalization, silent mouthing, silent reading.

35
Common brain regions for skilled word reading

• Encouragingly, despite these experimental
  differences, a set of brain regions common to
  single word reading has emerged (Fiez and
  Petersen 1998, Turkeltaub et al 2002, Palmer et
  al, in press)

12 published PET and fMRI studies
involving reading single words aloud adapted and
updated from Fiez and Petersen 1998
36
Digression What is the goal of functional
imaging?

• In contrast to a localist assumption of a
  one-to-one mapping between cortical regions and
  cognitive operations, an alternative view is that
  cognitive task performance is subserved by
  large-scale cortical networks that consist of
  spatially separate computational components, each
  with its own set of relative specializations,
  that collaborate extensively to accomplish
  cognitive functions. Carpenter et al 2001

37
Localization of Cognitive Operations in the Human
BrainPosner, Petersen, Fox, and Raichle 1988
Science

• The hypothesis is that elementary operations
  forming the basis of cognitive analyses of human
  tasks are strictly localized. Many such local
  operations are involved in any cognitive task. A
  set of distributed brain areas must be
  orchestrated in the performance of even simple
  cognitive tasks. The task itself is not
  performed by any single area of the brain, but
  the operations that underlie the performance are
  strictly localized

38
Localization of Cognitive Operations in the Human
BrainPosner, Petersen, Fox, and Raichle 1988
Science

• This form of localization of function differs
  from the idea that cognitive tasks are performed
  by a particular brain area. Visual imagery, word
  reading, and even shifting visual attention from
  one location to another are not performed by any
  single brain area. Each of them involves a large
  number of component computations that must be
  orchestrated to perform the cognitive task.

39
What is an area? How about a region?

• A neocortical area is defined by its afferents,
  efferents, architecture (cyto-, chemo-, myelo-)
  and function (e.g. primary motor cortex versus
  primary somatosensory cortex).
• FMRI does not (necessarily) show activation in
  areas.
• Check out The anatomical basis of functional
  localization in the cortex by Passingham et al,
  Nat Rev Neuro 2002

40
What does diffuse/distributed look like?
?
?
41
Methodological Issues in Studying the Development
of Reading with fMRI
42
Perceived Barriers

• Variability of child brain
• too variable to be compared directly with the
  adult brain.
• Performance mismatch on cognitive tasks children
  will not perform as well as adults on most tasks.
  
• performance versus processing
• These issues are relevant to any group-wise
  comparison
• Adults versus Children
• Princeton versus Yale

43
Issues

• Anatomical variability across development
• Physiological variability across development
• Performance differences between adults and
  children
• Task B problem
• choosing appropriate comparison tasks

44
Strategy

• Child-friendly (yet adult-challenging) tasks
• lexical processing tasks with overt responding
• Event-related design
• Relate performance to fMRI measures on
  trial-by-trial basis.
• Code and analyze only correct responses.
• Compatible with overt verbal responding.

45
Strategy

• Child-friendly (yet adult-challenging) tasks
• lexical processing tasks with overt responding
• Event-related design
• Relate performance to fMRI measures on
  trial-by-trial basis.
• Code and analyze only correct responses.
• Compatible with overt verbal responding.

• Voxel- and Region-wise (ANOVA) direct statistical
  comparison in a common stereotactic space
• Main effect of time image
• Group x time interaction image
• Performance-matching
• Maturation versus performance
• In scanner behavioral data

46
Transformation of adult and pediatric brains into
a common stereotactic space

• 1. Anatomical Variability
• 2. Functional Variability

47
Comparison of primary sulcus location and general
brain shape in children and adultsBurgund et al
Neuroimage 2002

• 20 adults, 20 children ages 7 and 8
• Brains placed in adult stereotactic space
• Measured
• parts of 10 sulci, distributed across the
  cortex in both hemispheres
• outer surface in transverse, sagittal and
  coronal planes at selected slice locations

48
There are only small differences between adults
and children in sulcus location
Burgund et al Neuroimage 2002
49
This is true for variability as well
Burgund et al Neuroimage 2002
50
Transformation of adult and pediatric brains into
a common stereotactic space

• 1. Anatomical Variability
• 2. Functional Variability

51
The basis of fMRI is the BOLD Effect (blood
oxygen level dependent)
BOLD Hemodynamic Response Function

• deoxyHg paramagnetic oxyHg not
• Neural activity-gtluxury perfusion
• deoxyHg/ oxyHg decreases
• Given volume less magnetic
• T2 relaxation in given volume slows
• T2 change is the BOLD effect

TIME (SEC)
0
4
14
32
From Josephs, Turner, Friston 1997
52
Methods

• Subjects
• 16 7 and 8 year-old children ages (8 male mean
  age 8.1)
• 16 adults (8 male mean age 26.4)
• Task
• Subjects pressed a button at the onset and offset
  of a foveal (3) flickering checkerboard. Both
  right and left index fingers were used for onset
  and offset, across subjects.

Kang et al, Neuroimage 2003
53
MANOVA showed that locations of foci were not
statistically different between children and
adults, in bilateral sensorimotor cortex and
bilateral SMA
Mean coordinates ?S.D.
Kang et al, Neuroimage 2003
54
ANOVA (age x time) interaction revealed no
timecourse differences except in right SM medial
region.
Kang et al, Neuroimage 2003
55
Anatomical and functional variability exists, but
issmall after transformation of pediatric and
adult brains into the same stereotactic space.

• Small is in reference to the spatial resolution
  of fMRI data (6-7mm)
• Offset is rarely greater than 4 mm. When
  variability is less than 5 mm, the likelihood of
  false-positive functional differences is very
  low.
• Hence, the degree to which post-transformed
  brains differ between children and adults, by
  this measurement, is beneath the resolution of
  fMRI.

56
Behavioral Data in the Scannercritical for
developmental studies

• Relate fMRI measures to concurrent behavior
• Assure compliance with task
• Acquire performance data
• Accuracy and reaction time
• Contend with performance confound
• Performance can be discrepant for the comparison
  task, as well
• Relevant to the Task B problem

57
Digression Task B problem

• Very difficult methodological issue
• Relevant to any group-wise comparison

58
Choosing appropriate comparison tasks

• Assumption Task A - Task B Activity of
  interest
• (Task Achild - Task Bchild) vs (Task Aadult -
  Task Badult)
• Interpreted as Task Achild vs Task Aadult
• Assumed that Task Bchild Task Badult
• But Task Bchild vs Task Badult is rarely
  presented

59
The Task B Problem
Assume (TaskB 1 TaskB 2)
Then (TaskA1 gt TaskA2)
60
The Task B Problem
But, if (TaskB 1 TaskB 2), And (TaskA1 -
TaskB1) 0, And (TaskA2 - TaskB 2) 0,
Then (TaskA1 TaskA2), and the image does not
follow from the assumptions
61
The Task B Problem
Alternatively, TaskB 1 lt TaskB 2, And TaskA1
TaskA2 TaskB 1 (TaskA1 - TaskB 1) 0 (TaskA2
- TaskB 2) negative
Then (TaskA1 - TaskB 1) - (TaskA2 - TaskB 2) is
positive because of the Task B
difference design cannot rule out this possiblity
62
The Task B Problem
Or, TaskA2 lt TaskB 2, And TaskA1 TaskB 1
TaskB 2 (TaskA1 - TaskB 1) 0 (TaskA2 - TaskB
2) negative
Then (TaskA1 - TaskB 1) - (TaskA2 - TaskB 2) is
positive driven by the relatively low activity
of TaskA2
63
Lexical Task Conditions

• Passive presentation
• Simple
• Single word reading
• Single word repetition
• Controlled
• Verb Generate nose --gt smell
• Opposite Generate good --gt bad
• Rhyme Generate book --gt look
• both auditory and visual modalities

64
Trial Design Visual Stimuli Run

• One task and one modality per trial
• Vocabulary open class words from lists of first
  200 reading words
• Performance responses recorded, scored, and
  reaction times measured (Soundedit)

65
Main effect of time image
Shows regions that have enough statistical
reliability of time course across groups to show
a main effect
66
Main effect of time image
Left Hemisphere Reading 21 adults 24 11-17
year olds 30 7-10 year olds
67
Main effect of time image
68
Interaction of group and time
Shows regions that have statistically reliable
differences in the time courses of the two groups
69
Interaction of group and time
Left Hemisphere Reading 21 adults 24 11-17
year olds 30 7-10 year olds
70
Interaction of group and time
71
Observations from the dataset
Qualitative comparison of comparisons of
repetition, reading and controlled tasks in
adults and young children Quantitative
comparison of simple and controlled tasks across
age range
72
Observations from this dataset
Qualitative comparison of comparisons of
repetition, reading and controlled tasks in
adults and young children Quantitative
comparison of simple and controlled tasks across
age range
73
Qualitative comparison of comparisons
Repetition - 7-10 year olds should be relatively
adult-like and thus interaction images should
show relatively small differences Controlled
tasks - Adults might have more fluent access to
metalinguistic capabilities than 7-10 year olds
and interactions may show more
differences Reading - Adults should be more
skilled maybe intermediate level of differences
74
Interaction of group and time for READING
75
Interaction of group and time for REPETITION
76
Interaction of group and time for CONTROLLED
TASKS
77
Interaction of group and time across
reading, repetition and controlled tasks
Repetition - Comparison shows relatively little
difference in functional anatomy Controlled
tasks Comparison shows more differences Reading
- Comparison shows intermediate level of
differences
Qualitative hypothesis seems pretty good
78
Major issue in comparing children to
adults Performance differences can create
confounds

• Task cond. Adult RT () Child RT ()
• Simple Read 665 (100) 668 (100)
• Controlled Read 1471 (86) 1880 (75)

79
Major issue in comparing children to
adults Performance differences can create
confounds

• Task cond. Adult RT () Child RT ()
• Simple Read 665 (100) 668 (100)
• Controlled Read 1471 (86) 1880 (75)

80
Major issue in comparing children to
adults Performance differences can create
confounds

• Task cond. Adult RT () Child RT ()
• Simple Read 665 (100) 668 (100)
• Controlled Read 1471 (86) 1880 (75)

Is the difference in the Controlled/interaction
images due to difficulty or age?
81
One way of dealing with performance differences
Performance assessment
82
Three developmental functional relationships Schla
ggar et al Science 2002
Age- Related
Performance- Related
Age/Performance Independent
p lt 0.003
p lt 0.001
p 0.600
Children/Adults Before Matching
p 0.450
p lt 0.014
p 0.470
Children/Adults Performance Matched
83
Different regions show different age-related
profiles
84
(No Transcript)
85
60 age-related regions of interest
86
Maturation of 60 age-related regions
87
Implications for functional brain development

• NOT strictly progressive,
• where dormant regions turn on

• NOT purely regressive,
• where active regions scale back

A combination of mechanisms
88
Some Interim Conclusions

• Developmental studies in children 7 years of age
  and up appear to be tractable using methods
  similar to those used in adults.
• Many changes in functional anatomy from 7 to
  adulthood do not appear to be related to simple
  performance variables. These changes seem to be
  more extensive in tasks with more complex
  demands.
• These changes do not seem to relate to a single
  type of mechanism, but rather reflect both
  progressive and regressive phenomena.

89
Some Interim Conclusions

• A strategy incorporating
• direct statistical comparison
• transformation to a common stereotactic target
  atlas
• event-related design with overt verbal responding
• performance matching
• can be implemented successfully to study
  cognitive development from age 7 to adulthood.

90
A handful of recent developmental fMRI studies of
reading
91
Disruption of posterior brain systems for reading
in children with developmental dyslexia.
Shaywitz et al 2002

• Goal to differentially tap the component
  processes in normal and impaired reading.
• TASKS
• Line / / / \ - / \ / / (visuo-spatial)
• Case t - V (letter identification)
• Single Letter Rhyme t vs v (sounding out
  letters)
• Non-word Rhyme jete vs leat (sounding out
  non-words)
• Semantic Category Judgement rice vs corn
  (semantics)

92
Shaywitz et al study

• 144 right handed children (70 dyslexic), m-13.3
  years for dys, 10.9 for normals.
• 1.5 T magnet
• Standard EPI imaging
• Standardized anatomical space
• Generate image for each contrast for each subject
  and then generate composite image

93
NWR-Line CAT-Line
94
Positive correlation between reading skill (Word
Attack) and activation across all subjects
(normal and dyslexics).Highlight the left
occipito-temporal cortex
95
Positive correlation between age and activation
in normal readers left inferior frontal cortex
for the CAT taskNo correlations mentioned in
temporoparietal nor occipitotemporal cortex.
96

• Children disengage posterior right hemisphere
  visual representations that interfere with proper
  word identification (which is in posterior left
  hemisphere). Orton 1925
• Children preferentially engage the dorsal
  (temporoparietal) decoding system and then
  transition to the ventral word form area with
  reading expertise. E.g. Pugh, Shaywitz

97
How do the neural systems responsible for reading
change throughout the period of its acquisition?

• Implicit word-processing task
• although subjects are not instructed to read the
  words, reading occurs obligatorily without
  conscious effort, resulting in comparable brain
  activity to that associated with explicit reading
  tasks.
• In adults
• Even novice readers can perform the task
  accurately because subjects are not explicitly
  required to read the words
• Explicit vs implicit task performance

98
Methods

• 57 right handed subjects (ages 6-22)
• 16 excluded for various reasons
• Neuropsychology battery
• Correlate brain activity with measures in this
  battery.
• Button press with right hand ascenders, left
  hand not

99
Results

• Although task performance was related to age

100
Results

• accuracy and RT differences between words and
  false font strings were not related to age
• Accuracy and RT differences were not related to
  reading ability

101
Results

• Post-test forced choice recognition to confirm
  implicit processing of stimuli
• 65 words
• 51 false fonts
• Age effects ?
• Word recognition accuracy correlated with the
  letter/word identification subtest (single word
  reading task)
• Strong correspondence between implicit word
  processing and reading ability

102
Implicit Reading

• Contrasts
• Words vs fixation
• False fonts vs fixation
• Words versus false fonts
• the contrast of words versus false font strings
  revealed those structures engaged by the implicit
  processing of words

103
Reading Acquisition

• Voxel wise regression between word-false font
  image and reading ability
• Composite score
• Reading of single words
• Novel word decoding
• Passage reading rate and accuracy
• Positive correlations with left hemisphere
  regions
• Negative correlations with right hemisphere
  regions

104
Reading Acquisition

• No relationship between reading ability and
  activity in the word form area.
• Development of ventral extrastriate via
  disengagement of right sided regions

105
Age-related changes in single word reading

• One hundred eleven right-handed subjects (ages 7
  to 35)
• single word reading
• event-related fMRI
• overt verbal responses
• A subset of 75 subjects with well-matched
  performance
• accuracy 100, and reaction time 665ms
• grouped by age (1) 7-10 yrs (n30) (2) 11-17
  yrs (n24) (3) 19-35 yrs (n21).
• Voxel and region-wise ANOVA were implemented to
  identify age-related regions.

106
Read Left Hemisphere
107
Read Right Hemisphere
108
Age-related changes in single word reading

• The majority of age-related regions showed
  decreases in activation across maturation.
• relatively greater activation in the youngest
  subjects transitioning to little or no activation
  in adults
• bilateral precuneus posterior cingulate, left
  angular gyrus
• relatively little (de)activation in the youngest
  subjects transitioning to robust deactivation in
  the older groups
• anterior cingulate, left caudate.
• Decreasing regions reached mature levels of
  activation in the 11-17 year old group (based on
  post hoc ANOVA).

109
Age-related changes in single word reading

• In the context of prior findings, these results
  demonstrate that the functional anatomy of simple
  lexical processing, as for controlled tasks,
  differs across development, independent of task
  performance.

110
Simple Read Interaction of group and time
111
Simple Read Interaction of group and time
112
Some Take Home Messages

• identical performance can be supported by
  non-identical functional neuroanatomy.
• The Strategy can help alleviate some of the
  methodological issues in studying development
• And group comparison, per se

113
B.L.S. is a Scholar of the Child Health Research
Center of Excellence in Developmental Biology at
Washington University School of Medicine
(HD33688).
Acknowledgements
Steven E Petersen PhD Tim Brown BS Kristina
Visscher BS Kristin Wenger BS Erica Palmer
MS Christine Kang MD Darcy Burgund PhD Jim Kelly
BS Randy Buckner PhD Avi Snyder MD PhD David Van
Essen PhD
Fran Miezin MS Mark McAvoy PhD Heather Lugar
BS Becky Coalson BS Tara Spevack PhD
NIH NINDS NICHDMcDonnell Center for Higher
Brain Function Charles A Dana
Foundation Burroughs-Wellcome Fund John Merck
Scholars Fund
114
Repeat Left hemisphere
115
Repeat Right Hemisphere
116
Repeat vs Read
117
Regions 17 and 4
Region 4
Region 17
118
A Priori Region of Interest
119
Voxel Counting
120
fMRI identifies regional specialization of neural
networks for reading in young children Gaillard
et al 2003

• 16 normal right handed children (m 7.2 years)
• 1.5T magnet
• Boxcar design experimental reading task and
  control visual task (looking at dot patterns).
• Reading tasks were skill-adjusted
• Silent task unmonitored. Post task test for
  comprehension (pop-quiz)
• Collected neuropsychological data
• No adult comparison built in

121
Gaillard et al 2003

• Children showed most activation in the left
  midtemporal (implicated in semantic processing)
  and inferior temporal gyri (fusiform and lingual
  implicated in word form), left inferior and mid
  frontal gyrus (implicated in grammatic decoding,
  verbal working memory and speech planning), and
  SMA

122
Gaillard et al 2003

• Children showed most activation in the left
  midtemporal (implicated in semantic processing)
  and inferior temporal gyri (fusiform and lingual
  implicated in word form), left inferior and mid
  frontal gyrus (implicated in grammatic decoding,
  verbal working memory and speech planning, and SMA

123
Gaillard et al 2003

• The neural networks that process reading are
  strongly lateralized and regionally specific by
  age 6-7 years.
• Neural networks in early readers are similar to
  those in adults.