NIH MRI Study of Normal Brain Development

1
NIH MRI Study of Normal Brain Development

• AC Evans Ph.D.
• Brain Development Cooperative Group
• Pediatric Functional Neuroimaging
• a Trans-NIH Workshop
• May 25, 2004

2
Contrast changes over time
3
Problems with previous studies

• Sample sizes too small to detect subtle signals
• Heterogeneity of subject population
• Little longitudinal data
• Lack of demographic representativeness
• Limited behavioral data for brain-behaviour
  correlation
• Limited MRI data (typically T1 only)
• Usually limited analysis techniques

4
MRI Study of Normal Brain
Development(N500)

• Create a database of behavioral and brain MRI
  development data for 0-18 years
• Analyze structural-behavioural relationships
• Develop technique for dissemination of results

5
Rationale for Project Design

• Problem
• Existing normative databases limited in size
• Pooling of databases difficult. Existing
  databases incompatible in
• Slice thickness
• Pulse sequence
• Demographics
• Behavioural tests
• One centre cannot collect large dataset fast
  enough to keep pace with technology
• Solution
• Clinical trial model multi-centre acquisition,
  uniform protocol

6
MRI Objectives

• Objective 1 Anatomical MRI/Behavior (5-18)
• Objective 2 Anatomical MRI/Behaviour (0-4)
• Ancillary A MR Spectroscopy
• Ancillary B Diffusion Tensor Imaging, Relaxometry

7
Pediatric Study Centers (PSCs)

• Neuropsychiatric Institute and Hospital, UCLA
• McCracken
• Childrens Hospital, Boston
• Rivkin
• Childrens Hospital of Philadelphia
• Wang
• University of Texas-Houston Medical School
• Brandt
• Childrens Hospital Medical Center, Cincinnati
• Ball
• Washington University, St. Louis
• McKinstry

8
Data Coordinating Center (DCC)

• Overall Direction Evans
• Database Zijdenbos, Vins,
  Charlet, Harlap, Das
• Behavioral Liaison Leonard, Milovan
• MRI Acquisition Pike, Arnaoutelis
• MRI Analysis Collins, Kitching,
  Lerch
• Sampling Plan Lange (Harvard)
• Data Transfer Zeffiro, Van Meter
  (Georgetown)
• Scientific Liaison Paus
• Clinical Liaison Ad-Dabbagh, Webster

9
Clinical Coordinating Center (CCC) St.
Louis Recruitment, behavioral measures
selection/certification, exclusions etc. for Obj
1,2 Botteron, Almli Behavioral QC
Rainey, Henderson,
Singer, Smith, Dubois, Warren,
Edwards DTI Processing Center (DPC) -
NIH Pierpaoli, Basser, Rohde, Chang MRS
Processing Center (MPC) UCLA (?) Alger,
ONeill
10
NIH MRI Study of Normal Brain Development
DCC
CCC
DPC
11
Recruitment Procedure

• Representative sample based on US 2002 census
• Zip code demographic data
• Telephone brief screener at recruitment
• Telephone long screener for inclusion criteria
• DISC, FIGS, CBCL
• Hospital Visit (Neuro exam, Behaviour, MRI)
• Objective 1 scans 3 times, every 2 years
• Objective 2 scans 3-6 times
• SES (3 levels) X age (0-18 yrs) X gender X
  ethnicity

12
Accrual by Age (Objective 1)
13
Family Income (Raw)
N 409
14
Parental Education
N 409
NIH MRI Study of Normal Pediatric
Development US
Population
15
Ethnicity
N 409
16
Behavioral Maturation is multi-dimensional
17
Test Battery
18
Study Organization
Study
Subject n
Subject n1
Subject n-1
Visit 2
Visit n
Visit 1

Exclusionary Screening Behavioral Instruments MRI
Procedures
DPS4
CBCL
DISC
CBCL
WASI
WJ3
CANTAB
JTCI
MRI
MRI
19
System Architecture
20
New technology never works first time
21
Candidate Profile
are identified by
22
DBMS Software Platform

• MySQL DBMS
• Cross-platform, open source
• Robustness, speed, reliability
• Low development cost
• - Remote management
• Graphical User Interface
• Cross platform, Internet enabled application
• PHP-based application, complemented by
• JavaScript, Java, and Perl for data
• manipulation tasks.
• Remote management, customizable

23
Database GUI
24
(No Transcript)
25
Database Summary

• Low-cost, extensible, secure
• 61 tests, approx.
• 20,000 possible data fields (1000
  filled/subject)
• Laptop-based behavioral test battery
• Automatic MRI data transfer
• Web-based behavioral GUI
• Interactive 3D MRI web viewer
• Automatic QC procedures
• Project web site

26
N449
N56
N188
Obj 1
Obj 2
DTI
27
Behavioral Instrument Status
Total of Instruments 9827
Objective 1 9029 Objective 2
798 Total in Database 9827
As of May 20, 2004.
28
IQ Scores (n248)
FIQ 110.70.8
VIQ 109.908
PIQ 108.90.8
29
Summary of Behavioral QC
30
W-J Passage Comprehension (n278)
r0.85, p0.000
31
WASI Vocabulary (n248)
r0.86, p0.000
32
Spatial Working Memory (CANTAB) Errors (n250)
r-0.75, p0.000
33
Objective 1
34
Objective 1 MRS/I
Objective 1 DTI
Corrected to exclude the early Cincinnati and
St. Louis 1 subjects since the DTI product was
not available.
35
Objective 2
36
QC Overview

• Goal quick turn-around time
• mean time for expedited review 1.5 days
• median time over all subjects 7.0 days
• Concentrated mostly on subject QC

37
Inter-packet movement
After
Before

• Separate volume into packets
• Register each packet to target
• Resample and interpolate to 1mm slice thickness

38
Data Flow for Brain Mapping
Data Acquisition Reconstruction Conversion to
MINC Image Format
PET
fMRI
aMRI
Registration aMRI - ePET aMRI - tPET
Frame alignment
Inter-slice normalization
T1/T2/PD/ alignment
Registration aMRI - fMRI
Partial volume correction
Intensity non-uniformity correction
Voxel-based model fitting
Voxel-based coherence analysis
3D segmentation
Stereotaxic Spatial Normalization
Inter-volume normalization
GLM analysis in 3, 4, or 5D
39
Anatomical MRI analysis pipeline
ASP
40
INSECT
ANIMAL
SEAL
ASP
41
Objective 1 classification
42
Obj 1 Tissue SPAMs (n337)
43
Age-related changes in WM density
Paus et al, Science 1999 n111
NIHPD n204 16 1 0 t10.5
44
WM density and Spatial Working Memory Between
Errors (Age removed)
NIHPD n188 7 -1 66 t-4.0
45
Cortical Surface Extraction(Kim, Hanyang U.)
46
Cortical Surface Extraction(Kim, Hanyang U.)
47
Automated extraction of both cortical surfaces
using CLASP algorithm (5 different
brains)
48
Analysis of detection limits

• 19 T1 MRIs of the same subject (Colin Holmes)
  with 1mm isotropic sampling.
• Computation of standard deviations across cortex.
• Across blurring kernels.
• Across metrics.
• Power analysis
• N needed to recover change of 0.5 mm
• Change required at N25 in each group.

49
Colins 19 Brains
Average
50
Required N to recover 0.5 mm
Unblurred
5mm
10mm
20mm
200
0
51
Recoverable change when N25
2mm
0mm
52
Prefrontal atrophy in normal aging (N851)
0.025
Slope (mm loss)
0.01
53
Cortical thickness vs. age between (Obj 1, 4-18)
(slope in mm/yr, N289)
54
Cortical thickness versus Age
55
(No Transcript)
56
(No Transcript)
57
Obj 2 stereotaxic T1 average
nihpd120_obj2
icbm152
58
Related large-scale projects

• ICBM (7000)
• Giedd and Rapoport (3000)
• Brad Peterson (TS, OCD, ADHD 600)
• Maternal Adversity (MAVAN 500)
• Tourettes Neuroimaging Consortium (500)
• Japanese Human Brain Project (1200)

59
Welcome to the good ship NIHPD