Function BIRN: fMRI as a window to brain function

1
Function BIRN fMRI as a window to brain function

• Jessica Turner, Ph.D.
• University of California, Irvine

2
The FBIRN goal Develop multi-site neuroimaging
techniques
Are positive symptom schizophrenics associated
with more severe superior temporal gyrus
dysfunction?
Integrated View
Web
PubMed, Expasy, Brain Map, fMRI DC
fMRI
Clinical
ERP
Structure
3
FBIRN Multi-site imaging study
UMN
MGH
BWH
Stanford
Yale
UI
UCLA
UCI
Duke/ UNC
UNM
UCSD
4
Multi-site federated datasets
5
FMRI as a method

• Localization Where differences in neural
  activity in the brain took place

• This can be used to determine
• What areas are active when a person does a
  certain cognitive task or another
• How one group of subjects differs from another in
  cognitive response
• How a certain disease/treatment changes that
  cognitive activity

6
Schizophrenia as a model

• Schizophrenia
• Brain illness with subtle structural and
  functional changes
• Active area of imaging research with many
  competing theories and approaches

7
Functional MRI studies of schizophrenia

• What are the pathophysiologies of brain function
  underlying schizophrenia?
• Cortical dysfunctions revealed by fMRI in
• Working memory
• Recognition memory
• Verbal processing
• Auditory processing
• Face perception
• Motor task timing
• Basic visual perception
• Resting state circuitry

8
Schizophrenia and working memory

• (A) Left Statistical map showing bilateral
  dorsolateral prefrontal cortex (DLPFC) activation
  in an unmedicated schizophrenic subject during a
  working memory task
• (B) Significantly activated voxels are
  represented in red for the normal group, blue for
  the schizophrenia group, and yellow for both
  groups (overlap).

From Manoach, et al., 2000.
9
FMRI variance and correction
How big a problem is it?
10
Traveling Humans Study (Phase I)
HIPAA
HIPAA
HIPAA
HIPAA
HIPAA
Subjects traveled around the country to be
scanned at all FBIRN sites Unique dataset
Subject x site interactions can be measured for
the first time
11
Measuring Subject and Site Effects
12
Calibration methods

• Scaling by the breathhold response increases
  group effect size

Group Differences
Corrected
Uncorrected
Differences between young and old subjects in an
fMRI task, before and after correcting for BOLD
differences in a separate breathhold task
13
Calibration methods

• Smoothing to a common level reduces intersite
  effects

14
Multi-site clinical imaging study (Phase II)

• Will these calibration methods work in a
  multi-site imaging study on patients?
• Each FBIRN site collecting 15 schizophrenic
  subjects and 15 controls
• In a common imaging paradigm
• Using the same combination of calibration and
  cognitive tasks
• Includes the challenges of multi-site clinical
  populations

15
Controls N41, plt1x10-10
Patients N44, plt1x10-5
MFG, Right Hemisphere, Visit 1
Visit 1
Slice Depth Y-36
Controls Patients
Visit 2
Brain function in a given region can be driven by
several tasks
MFG, Right Hemisphere, Visit 2
16
Impact of Calibration Method
ANOVA Observed Effect Size
Cohens f
None Smooth
Smooth, BH Smooth, BH
Calibrate Calibrate, BH Screen
17
Tools delivered since FBIRN inception, 2003

• Data collection
• Phantoms MRI calibration and QA protocol
• E-prime programs used in data collection
• Data management
• Human Imaging Database schema/Graphical User
  Interface Code
• Clinical Assessment Layout Manager/General
  Assessment Manager Engine
• XCEDE XML schema
• Tools for reading and writing XCEDE XML image
  wrappers
• Standardized, general purpose directory hierarchy
• Datasets
• Phase I, 10-site traveling subjects dataset
• MRI QA datasets

18
FBIRN Neuroinformatics Federated Databases are
functioning!
p1
p2
UMN HID
p1
MGH HID
p2
BWH HID
p1
p2
Stanford HID
Yale HID
p1
UI HID
p1
p2
p2
UCLA HID
p2
UCI HID
p1
Duke HID
p1
p2
UNM HID
p1
p2
UCSD HID
p2
p1
19
Reproducibility Phantom QA methods
20
FBIRN Tool Sharing

• QA stability phantom methods for assessing image
  noise have been shared and adopted as standards
  for accepting new scanners
• .
• The FBIRN XML image description headers are
  helping to shape national standards of imaging
  data for neuroimaging and other fields, through
  coordination with NIfTI.
• The tools for neuroimaging visualization have
  been adapted to cardiac imaging to allow
  cardiologists to see frame by frame dynamic
  cardiac perfusion images.
• Other multi-site consortia are exploring our
  database schema as a datasharing mechanism.

21
Where to find us

• www.nbirn.net
• http//www.nbirn.net/Resources/Downloads/

22
(No Transcript)
23
FBIRN External Collaborations

• TURNS
• MIND
• NIfTI
• ADNI
• NA-MIC
• GCRCs
• Other users of the HID, FIPS, calibration
  methods, etc.

24
The future of multi-site imaging data integration

• New and better multi-site Calibration methods
• Using BH improvements
• Using ASL
• Broader multi-site Cognitive tasks
• Efficient, combined tasks
• Activating emotional and cognitive circuits, and
  interactions
• Robust multi-site Analysis methods and tools
• Publicly available, mediated datasets for queries
  and analysis
• Traveling subjects dataset now available

25
Phase II Study Image Data Volume

• 21,038 raw image files per subject
• 2.4 gbytes of raw image data per subject
• 25 gbytes to 40 gbytes of processed image data
  per subject (depending on hypotheses tested)
• 10 million slices of functional imaging data in
  Phase II
• 7 tbytes of image data for all of the Phase II
  analyses (conservative estimate of 25
  gbytes/subject)

FBIRN Shared Data Files
26
Generic Image Upload Scripts
Local File System
FMRI Images
HIDB

• Clinical Data
• Computer Aided Scale Input
• (CALM/GAME)
• Clinical Data Entry Interface

27
Generic Image Upload Scripts
Local File System
FMRI Images
HIDB

• Clinical Data
• Computer Aided Scale Input
• (CALM/GAME)
• Clinical Data Entry Interface

28
FBIRN Informatics Vision
FMRI Images Automated Image Upload to SRB/HID for
sharing
URL links
HIDB

• Clinical Data
• Computer Aided Scale Input
• (CALM/GAME)
• Clinical Data Entry Interface

29
Current FBIRN Informatics
FMRI Images Automated Image Upload to SRB/HID for
sharing
URL links
HIDB

• Clinical Data
• Computer Aided Scale Input
• (CALM/GAME)
• Clinical Data Entry Interface

30
fMRI as a method
Breathholding paradigm Block trials--15s on,
15s off, 8 blocks. On block Hold breath while
circle of diminishing size is projected. Off
Block Fixation cross, breathe normally
31
fMRI as a method

• Gray matter shows a change in signal across the
  brain.
• Example data from Stanford
• University from a breathhold
• task in one subject.

32
Current Datasets Phantom QA Data from all sites
on BIRN Portal
https//portal.nbirn.net/BIRN/cgi-bin/DataGrid/MRI
s/publicMRIStability.cgi
33
Interoperability XCEDE
INPUTS
Analyze7.5
DICOM
GE Pfile
Format X
add XML
add XML
add XML
add XML

• Readable by humans
• Expandable
• More efficient and less prone to errors

XML-Aware Applications (e.g., FIPS, BIRN QA)
PROCESSING STEP
FIPS Analysis, QA plots, etc.
OUTPUTS
34
FSL Image Processing Scripts (FIPS)

• Front end interface for automated image
  processing and analysis
• Open source

35
Calibration methods

• Calibrating BOLD signal by SFNR in grey matter
  reduces intersite effects

B
Intersite effects in 1.5T scanners in eight ROIs,
before and after correction
36
Reproducibility Phantom QA methods
Time Series Stability
37
The Future of Multi-site Data Integration
Are positive symptom schizophrenics associated
with more severe superior temporal gyrus
dysfunction?
Integrated View
Mediator
Wrapper
Wrapper
Web
Wrapper
Wrapper
Wrapper
Wrapper
PubMed, Expasy, Brain Map, fMRI DC
fMRI
Clinical
ERP
Receptor Density
Structure
38
Multi-site MRI Datasets Problems

• Measurement Noise
• Geometric Differences and instabilities
• RF Field Differences
• K-Space Trajectory
• Field Strength

• Registration/Alignment and subject motion
• Stimulus Presentation Differences
• Physiological Noise
• Subject-Specific Variations

39
Multi-site methodologies and tools

• Scanner calibration
• Paradigm issues and standards
• Clinical data collection standards
• Analysis of large datasets
• Data sharing
• Data formats, metadata and data location
• All of these are needed to achieve
• multi-site functional imaging studies

40
The FBIRN Aims

• Calibration
• Assess and correct for the major sources of
  multi-site variation
• Cognitive Tasks
• Develop robust protocols for multi-site fMRI
  studies
• Analysis
• Develop methods to analyze multi-site fMRI data
• Tools
• Develop a scalable IT application toolkit to
  support multi-site fMRI studies
• Dissemination
• Make the tools, procedures, and collected
  datasets widely available

41
Automated Quality Assurance (QA)

• To remove spikes that occur in image collection

• In the traveling subjects dataset, these occurred
  in .01 of the images
• Looking for them visually is not feasible

Difference from the mean image

• Automated spike detection and removal allows
  millions of images to be processed correctly

42
Schizophrenia as a model

• Schizophrenia
• Brain illness with subtle structural and
  functional changes
• Active area of imaging research with many
  competing theories and approaches
• Progress hampered by
• Inconsistent data lack of replications
• Noncomparable imaging techniques
• Small and diverse patient populations
• This applies to many diseases

43
Multi-site Methods Required for Disease Study

• Representative sample
• Private, public, VA
• SES
• Ethnicity, gender, comorbidities
• Requirement for drug registration
• Rapid recruitment
• Deeper understanding of subtle cause-effect
  relationships