Twenty Years of Functional MRI - brainmapping.org€¦ · Contrast Basics 1992 ... E. C. Wong, Magnetic resonance imaging of human brain function: principles, ... Interpretation Applications

Peter A. Bandettini, Ph.D.!!

Section on Functional Imaging Methods!

Laboratory of Brain and Cognition!http://fim.nimh.nih.gov!

&!Functional MRI Facility!

http://fmrif.nimh.nih.gov!!

[email protected]!

Twenty Years of Functional MRI!

(1990) Science, 250, 53-61.!angiography!

Gadolinium perfusion!

Diffusion!magnetization transfer!

metabolic imaging (NAA)!

NAA! choline!

creatine! lactate!

Five Key Factors For The !Emergence of Functional MRI!

1.  Magnetic properties of red blood cells !2.  Activation related hemodynamic changes!3.  Spatial scale of brain activation!4.  Echo Planar Imaging!5.  Prevalence of MRI scanners!



red blood cells

L. Pauling, C. D. Coryell, Proc.Natl. Acad. Sci. USA 22, 210-216, 1936.!

!

K.R. Thulborn, J. C. Waterton, et al., Biochim. Biophys. Acta. 714: 265-270, 1982.!

!

S. Ogawa, T. M. Lee, A. R. Kay, D. W. Tank, Proc. Natl. Acad. Sci. USA 87, 9868-9872, 1990.!

!

Turner, R., Lebihan, D., Moonen, C. T. W., Despres, D. & Frank, J. Magnetic Resonance in Medicine, 22, 159-166, 1991.!

oxygenated

deoxygenated

Magnetic Properties of Blood!

Blood R2 proportional to Oxygenation!R2 effect is bulk susceptibility and not dipole-dipole!

S. Ogawa, T.-M. Lee, A. S. Nayak, P. Glynn, Magn. Reson. Med, 14, 68-78 (1990) !

100% O2!

20% O2!

in vivo!

100% oxygenated blood!

0% oxygenated blood!

in vitro!

R. Turner, D. LeBihan, C.T.W. Moonen, D. Despres, J. Frank, Magn. Reson. Med, 22, 159-166 (1991) !

“BOLD contrast adds to…functional MRI methodologies that are likely to be complementary to PET imaging in the study of regional brain activity.” !

Ogawa, S., Lee, T. M., Kay, A. R. and Tank, D. W. (1990) Proceedings of the National Academy of Sciences of the United States of America, 87, 9868-9872.!



Cerebral Tissue Activation!

Local Vasodilatation!

Increase in Cerebral Blood!Flow and Volume!

Oxygen Delivery Exceeds!Metabolic Need!

Increase in Capillary and Venous Blood Oxygenation!

Decrease in Deoxy-hemoglobin!Deoxy-hemoglobin: paramagnetic!Oxy-hemoglobin: !diamagnetic!

Decrease in susceptibility-related!intravoxel dephasing!

Increase in T2 and T2*!

Local Signal Increase in T2 and T2* - weighted sequences!

task task



Brain Function!

visual

Language, reading, speech

motor sensory motor planning

Behavioral control, cognitive, socialization

memory auditory

1991!

http://www.thebrain.mcgill.ca!

Visual Cortex Organization!

Yacoub et al. PNAS 2008!Scalebar = 0.5 mm

Orientation Columns in Human V1 as Revealed by fMRI at 7T

Phase 0° 180°

Phase Map

Yacoub et al. PNAS 2008

Functional Neuroimaging Techniques!

Log Time (sec)!

Log

Size

(m

m)!

Non-invasive!

Invasive!



MRI vs. fMRI!MRI! fMRI!

one image!

many images !(e.g., every 2 sec for 5 mins)!

high resolution!(1 mm)!

…!

T2* decay!

EPI Readout Window!� 20 to 40 ms!

Single Shot Echo Planar Imaging (EPI)!

August, 1991

1991-1992"

1992-1999"

1976 P. Mansfield conceives of EPI!1989 EPI of humans emerges on a handful of scanners!

! 3 x 3 x 3-10 mm3!1989 ANMR retrofitted with GE scanners for EPI!1991 Home built head gradient coils perform EPI!1996 EPI is standard on clinical scanners!2000 Gradient performance continues to increase!2002 Parallel imaging allows for higher resolution EPI!2006 1.5 x 1.5 x 1.5 mm3 single shot EPI possible!2009 At 7T sub – mm single shot EPI for fMRI is possible!

Approximate EPI Timeline!



“fMRI” or “functional MRI”!

Scopus: Articles or Reviews Published per Year!

EPI is sold on standard clinical scanners!

The Beginnings!

(1990) Science, 250, 53-61.

angiography

Gadolinium perfusion

Diffusion magnetization transfer

metabolic imaging (NAA)

NAA choline

creatine lactate

Resting Active

Susceptibility Contrast agent bolus injection and time series collection of T2 - weighted images

The First Functional MRI Results!

Susceptibility Contrast agent bolus injection and time series collection of T2 - weighted images

The First Functional MRI Results!

K. K. Kwong, et al, (1992) �Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation.� Proc. Natl. Acad. Sci. USA. 89, 5675-5679. S. Ogawa, et al., (1992) �Intrinsic signal changes accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging.� Proc. Natl. Acad. Sci. USA. 89, 5951-5955. P. A. Bandettini, et al., (1992) �Time course EPI of human brain function during task activation.��Magn. Reson. Med 25, 390-397. Blamire, A. M., et al. (1992). �Dynamic mapping of the human visual cortex by high-speed magnetic resonance imaging.� Proc. Natl. Acad. Sci. USA 89: 11069-11073. Frahm, J., et al (1992) �Dynamic MR Imaging of Human Brain Oxygenation During Rest and Photic-Stimulation.� Journal of Magnetic Resonance Imaging, 2, 501-505.

K. K. Kwong, et al, (1992) �Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation.� Proc. Natl. Acad. Sci. USA. 89, 5675-5679.

S. Ogawa, et al., (1992) �Intrinsic signal changes accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging.� Proc. Natl. Acad. Sci. USA. 89, 5951-5955.

P. A. Bandettini, et al., (1992) �Time course EPI of human brain function during task activation.� Magn. Reson. Med 25, 390-397.

2.5 cm !!

TR = 2 sec!TE = 50 ms!One slice!In plane 3.75 x 3.75!

1991

Blamire, A. M., et al. (1992). �Dynamic mapping of the human visual cortex by high-speed magnetic resonance imaging.� Proc. Natl. Acad. Sci. USA 89: 11069-11073.

Contrast Basics 1992…Perfusion using Arterial Spin Labeling

EPISTAR

FAIR

. . .

. . . Perfusion Time Series

��

Perfusion Contrast FAIR EPISTAR TI (ms)

200

400

600

800

1000

1200

Williams, D. S., Detre, J. A., Leigh, J. S. & Koretsky, A. S. (1992) “Magnetic resonance imaging of perfusion using spin-inversion of arterial water.” Proc. Natl. Acad. Sci. USA 89, 212-216.

Edelman, R., Siewert, B. & Darby, D. (1994) “Qualitative mapping of cerebral blood flow and functional

localization with echo planar MR imaging ans signal targeting with alternating radiofrequency (EPISTAR).” Radiology 192, 1-8.

Kim, S.-G. (1995) “Quantification of relative cerebral blood flow change by flow-sensitive alternating

inversion recovery (FAIR) technique: application to functional mapping.” Magn. Reson. Med. 34, 293-301.

Kwong, K. K. et al. (1995) “MR perfusion studies with T1-weighted echo planar imaging.”Magn. Reson.

Med. 34,878-887.

BOLD" Rest ! !Activation"

P. A. Bandettini, E. C. Wong, Magnetic resonance imaging of human brain function: principles, practicalities, and possibilities, in "Neurosurgery Clinics of North America: Functional Imaging" (M. Haglund, Ed.), p.345-371, W. B. Saunders Co., 1997.

Perfusion"

•  Volume (gadolinium) •  BOLD •  Perfusion (ASL) •  !CMRO2 •  !Volume (VASO) •  Neuronal Currents •  Diffusion coefficient •  Temperature

Functional Contrast Methodology

Interpretation Applications

Technology Coil arrays High field strength High resolution Novel functional contrast

Functional Connectivity Multi-modal integration Pattern-effect imaging Real time feedback Task design

Fluctuations Dynamics Spatial patterns

Healthy Brain Organization Clinical Pathology

J. Illes, M. P. Kirschen, J. D. E. Gabrielli, Nature Neuroscience, 6 (3)m p.205!

Motor (black)!Primary Sensory (red)!Integrative Sensory (violet)!Basic Cognition (green)!High-Order Cognition (yellow)!Emotion (blue)! 36 02 01 00 99 98 97 96 95 94 93 92 91 90 89 88 82

Methodology

Hemoglobin

Blood T2

IVIM

Baseline Volume

Interpretation

Applications

�Volume-V1

BOLD

Correlation Analysis

Linear Regression Event-related

BOLD -V1, M1, A1

TE dep

Veins

IV vs EV BOLD models

ASL

Deconvolution

Phase Mapping

V1, V2..mapping

Language Memory

Presurgical Attention

PSF of BOLD Pre-undershoot

Ocular Dominance

Mental Chronometry

Electrophys. correlation

1.5T,3T, 4T 7T

SE vs. GE

Performance prediction

Emotion

Real time fMRI

Balloon Model

Post-undershoot

Inflow

PET correlation

CO2 effect

CO2 Calibration

Drug effects

Optical Im. Correlation

Imagery

Clinical Populations

Plasticity

Complex motor

Motor learning

Venography

Face recognition

Children

Simultaneous ASL and BOLD

Surface Mapping

Linearity

Mg+

Dynamic IV volume

Bo dep.

Diff. tensor

Volume - Stroke

Z-shim

Free-behavior Designs

Extended Stim.

Local Human Head Gradient Coils

NIRS Correlation

SENSE

Baseline Susceptibility

Metab. Correlation

Fluctuations

Priming/Learning

Resolution Dep.

Tumor vasc.

Technology EPI on Clin. Syst. EPI

Quant. ASL

Multi-shot fMRI

Parametric Design

Current Imaging?

Multi-Modal Mapping

Nav. pulses

Motion Correction

MRI Spiral EPI

ASL vs. BOLD

>8 channels

Multi-variate Mapping

ICA

Fuzzy Clustering

Excite and Inhibit

03

Mirror neurons

Layer spec. latency

Latency and Width Mod

�vaso�

How fMRI Is Currently Used

Research Applications -map networks involved with specific behavior, stimulus, or performance -characterize changes over time (seconds to years) -determine correlates of behavior (response accuracy, etc…) -characterization of groups or individuals

Clinical Research

-clinical population characterization (probe task or resting state) -assessment of recovery and plasticity -attempts to characterize (classify) individuals

Clinical Applications

-presurgical mapping (CPT code in place as of Jan, 2007)

Top 10 Most Significant Developments in fMRI!

1.  “Resting state” correlations!

2.  SPM, FSL, AFNI, Brain Voyager processing platforms!

3.  Fast even related designs with deconvolution!

4.  Multi-modal integration / comparison (animal and human)!

5.  Hemodynamic response models!

6.  High field / high resolution results!

7.  Parallel acquisition / reconstruction (SENSE/SMASH)!

8.  Detailed retinotopic mapping!

9.  FMRI Decoding / Multivariate / Machine Learning!

10.  Robust motion correction or ICA analysis or real time fMRI or fMRI-adaptation !

Overview of fMRI

Functional Contrast: Blood volume Blood flow/perfusion Blood oxygenation

Spatial resolution:

Typical: 3 mm3 Upper: 0.5 mm3

Temporal resolution:

Minimum duration: < 16 ms Minimum onset diff: 100 ms to 2 sec

Sensitivity: tSNR = 40/1 to 120/1 fCNR = 1/1 to 6/1

Interpretability issues: Neurovascular coupling, vascular sampling, blood, physiologic noise, motion and other artifacts, etc..

task task

What fMRI Can�t Do What some would argue are shortcomings with fMRI

• Too low SNR vs subject/patient limits of compliance (about 2 hours)

• Requires motivated subjects/patients (motion sensitivity)

• Too low spatial resolution (each voxel has several million neurons)

• Any higher resolution than 3 mm3 lost with subject averaging.

• Too low temporal resolution (hemodynamics are variable and sluggish)

• Too inconsistent activation patterns

• Anatomical images for fMRI are low quality (dropout/distortion)

• Requires a task (BOLD cannot look at baseline maps).

• Too confined space and high acoustic noise (environment non-optimal).

• Too many physiologic variables influence signal.

What fMRI Can�t Do What some would argue are shortcomings with fMRI

• Too low SNR vs subject/patient limits of compliance (about 2 hours) • Higher field strength, multi-channel, temporal processing (15 min) • Requires motivated subjects/patients (motion sensitivity) • Real time fMRI, improved motion correction, navigator pulses • Too low spatial resolution (each voxel has several million neurons) • FMRI-adaptation paradigms, smallest functional unit about 0.5mm anyway • Any higher resolution than 3 mm3 lost with subject averaging. • Spatial averaging becoming a bit less common. • Too low temporal resolution (hemodynamics are variable and sluggish) • Paired pulse paradigms, timing modulation w/ latency comparison (<100 ms) • Too inconsistent activation patterns • Not typically due to low SNR. Subject differences – good problem. • Anatomical images for fMRI are low quality (dropout/distortion) • SENSE, multi-channel, and Bo correction can help. • Requires a task (BOLD cannot look at baseline maps). • Calibration methods getting much better. Baseline info is coming. • Too confined space and high acoustic noise (environment non-optimal). • Strategies around this. Vendors learning how to dampen sound. • Too many physiologic variables influence signal. • Again, calibration methods are improving

Exciting Trends!

• FMRI Decoding / Multivariate Analysis / Machine Learning!

• Resting State Fluctuations (Connectivity)!!• High Field / High Temporal & Spatial Resolution (input/output layers)!!• Longitudinal fMRI / MRI studies across time scales!

• Clinical inroads (default mode, genetic correlations, connectivity)!

• Multi-modal integration!

• Individual assessment!

• Multi-band sequences…short TR whole brain connectivity.!

1991

Parametric manipulation of brain activation demonstrated that BOLD contrast approximately followed the level of brain activation: visual system (Kwong et al., 1992), auditory system (Binder et al., 1994), and motor system (Rao et al., 1996). The use of continuous variation of visual stimuli parameters as a function of time was proven a powerful method for fMRI-based retinotopy: (Engel et al., 1994, Deyoe et al., 1994, Sereno et al., 1995). Event-related fMRI was first demonstrated (Blamire et al., 1992). Application of event-related fMRI to cognitive activation was shown (Buckner et al., 1996, McCarthy et al., 1997). Development of mixed event-related and block designs was put forward: (Donaldson et al., 2002). Paradigms were demonstrated in which the activation timing of multiple brain systems timing was orthogonal, allowing multiple conditions to be cleanly extracted from a single run (Courtney et al., 1997). High resolution maps were created: For spatial resolution: ocular dominance columns (Menon et al., 1997, Cheng et al., 2001) and cortical layer activation maps were created (Logothetis et al., 2002). Extraction of information at high spatial frequencies within regions of activation was demonstrated (Haxby et al., 2001). For temporal resolution: Timings from ms to hundreds of ms were extracted (Ogawa et al., 2000, Menon et al., 1998, Henson et al., 2002, Bellgowan et al., 2003). The development of “deconvolution” methods allowed for rapid presentation of stimuli (Dale and Buckner, 1997). Early BOLD contrast models were put forward: (Ogawa et al., 1993, Buxton and Frank, 1997). More sophisticated models were published that more fully integrated the latest data on hemodynamic and metabolic changes (Buxton et al., 2004). The development of “clustered volume” acquisition was put forth as a method to avoid scanner noise artifacts: (Edmister et al., 1999). The findings of functionally related resting state correlations: (Biswal et al., 1995) and regions that consistently show deactivation (Binder et al., 1999, Raichle et al., 2001) were described. Observation of the pre-undershoot in fMRI (Hennig et al., 1997, Menon et al., 1995, Hu et al., 1997) and correlation with optical imaging was reported (Malonek and Grinvald, 1996). Simultaneous use of fMRI and direct electrophysiological recording in non-human primate brain during visual stimulation elucidated the relationship between fMRI and BOLD contrast. (Logothetis et al., 2001). Simultaneous electrophysiological recordings in animal models revealed a correlation between negative signal changes and decreased neuronal activity (Shmuel et al., 2002). Simultaneous electrophysiological recordings in animal models provided evidence that inhibitory input could cause an increase in cerebral blood flow (Matheiesen et al., 1998). Structural equation modeling was developed in the context of fMRI time series analysis: (Buchel and Friston, 1998).

How most fMRI studies are performed

MRI parameters: 1.5T - 3T, 64 x 64 matrix, 3mm x 3mm x 5mm voxel size, whole brain, TR = 2 sec. Paradigm: Block design or event-related, single or multiple conditions. Analysis: Motion correct, multi-regression, spatial smoothing and spatial normalization, standard classical statistical tests, multi-subject averaging. Hypothesis: A region or network of regions show modulation with a task. This modulation is unique to the task and/or population.

How fMRI might be be performed MRI parameters: 3T - 11.7T, 256 x 256 matrix, 0.5 x 0.5 x 0.5 voxel size, whole brain TR = 1sec or select slab TR = 100 ms. Paradigm: Natural, continuous, fMRI-adaptation, or no stimuli/task. Simultaneous multi-modal, or multiple contrast measurements. Analysis: Motion correct, dynamic Bo-field correction, no spatial or temporal smoothing, machine learning algorithms, pattern classification, resting state connectivity assessment, hemodynamic parameter assessment - calibration, correlation with behavior. Hypothesis: Similar to previous but using the high resolution patterns, fluctuations, dynamics, and contrast mechanisms that we are still figuring out how to interpret and extract.

What fMRI Might Do

Clinical Research Complementarity -usage of clinical research findings for more effective diagnoses, prediction, characterization, and/or intervention

Clinical treatment and assessment of therapy -better understanding of the specific pathology mechanism -drug effect assessment -assessment of therapy progress, biofeedback -epileptic foci mapping -neurovascular physiology assessment

Non clinical uses

-lie detection -prediction of behavior tendencies -brain/computer interface

Neuronal Activation

Hemodynamics ? ? ?

Measured Signal

Noise

?

Interpretation Issues

Scale bar = 0.5 mm


Phase 0° 180°

Phase Map

Yacoub, Ugurbil & Harel

Real time fMRI feedback from Anterior Cingulate Cortex to reduce chronic pain

Control over brain activation and pain learned by using real-time functional MRI, R. C. deCharms, et al. PNAS, 102; 18626-18631 (2005)

SC NL KB

JL HG EE

CC BK BB

Individual activations from the left hemisphere of the 9 subjects

group Individual Differences in Brain Activations During Episodic Retrieval

Miller et al., 2002

Courtesy, Mike Miler, UC Santa Barbara and Jack Van Horn, fMRI Data Center, Dartmouth University

group Individual Differences in Brain Activations During Episodic Retrieval

Miller et al., 2002

KB NL SC

HG JL

BB BK CC

EE

Individual activations from the right hemisphere of the 9 subjects


Group Analysis of Episodic Retrieval

Subject SC

Subject SC 6 months later

These individual patterns of activations are stable over time


Finger Movement

Left Right

Finger Movement

Toe Movement

Left Right Left Right

Tactile Stimulation Finger Movement Tapping vs Pure Somatosensory

Simple Left Complex Left

Simple Right Complex Right Imagined

Complex Right

Imagined Complex Left

Reading

Listening to Spoken Words

≈ 5 to 30 ms"Pruessmann, et al. !

3T single-shot SENSE EPI using 16 channels: 1.25x1.25x2mm

Menon, R. S., S. Ogawa, et al. (1997). J Neurophysiol 77(5): 2780-7.

R. D. Frostig et. al, PNAS 87: 6082-6086, (1990).

Ocular Dominance Column Mapping

Optical Imaging

Cheng, et al. (2001) Neuron,32:359-374

0.47 x 0.47 in plane resolution



Technology


Scalebar = 0.5 mm


Phase 0° 180°

Phase Map


Visual Activation Paradigm: 1 , 2, & 3 Trials""

20 sec"0 sec"

0 sec"2 sec" 20 sec"

0 sec"2 sec" 20 sec"4 sec"

-"1"

0"

1"

2"

3"

4"

5"

0" 1" 2" 3" 4" 5" 6" 7" 8" 9"1"0"1"1"1"2"1"3"1"4"1"5"1"6"1"7"1"8"1"9"T"I"M"E" "("S"E"C")"

O N E - T R I A L

T W O - T R I A L

T H R E E - T R I A L

-"1"

0"

1"

2"

3"

4"

5"

0" 1" 2" 3" 4" 5" 6" 7" 8" 9"1"0"1"1"1"2"1"3"1"4"1"5"1"6"1"7"1"8"1"9"T"I"M"E" "("S"E"C")"

RAW DATA" ESTIMATED RESPONSES"

Rest: seed voxel in motor cortex

Activation: correlation with reference function

B. Biswal et al., MRM, 34:537 (1995)

Resting State Correlations

94!

Resting state networks identified with ICA

M. DeLuca, C.F. Beckmann, N. De Stefano, P.M. Matthews, S.M. Smith, fMRI resting state networks define distinct modes of long-distance interactions in the human brain. NeuroImage, 29, 1359-1367

Goldman, et al (2002), Neuroreport

BOLD correlated with 10 Hz power during “Rest”

Positive

Negative

10 Hz power

Fluctuations Patterns

1991

Ventral temporal category representations

Object categories are associated with distributed representations in ventral temporal cortex

Haxby et al. 2001

Pattern-recognition analysis of fMRI activity patterns

•  Haxby et al. (2001) •  Cox & Savoy (2003) •  Carlson et al. (2003) •  Kamitani & Tong (2005) •  Haynes & Rees (2005) •  Kriegeskorte et al (2006)

S.A. Engel, et al. Investigative Ophthalmology & Visual Science 35 (1994) 1977-1977.

PSF FWHM = 3.5mm

Logothetis et al. (2001) “Neurophysiological investigation of the basis of the fMRI signal” Nature, 412, 150-157

16 channel parallel receiver coil!

GE birdcage GE 8 channel coil Nova 8 channel coil

8 channel parallel receiver coil

J. Bodurka, et al, Magnetic Resonance in Medicine 51 (2004) 165-171.

≈ 5 to 30 ms"Pruessmann, et al. !

3T single-shot SENSE EPI using 16 channels: 1.25x1.25x2mm

Menon, R. S., S. Ogawa, et al. (1997). J Neurophysiol 77(5): 2780-7.

R. D. Frostig et. al, PNAS 87: 6082-6086, (1990).

Ocular Dominance Column Mapping

Optical Imaging





Technology

Yacoub et al. PNAS 2008 Scalebar = 0.5 mm


Phase 0° 180°

Phase Map


Documents

Twenty Years of Functional MRI - brainmapping.org€¦ · Contrast Basics 1992 ... E. C. Wong, Magnetic resonance imaging of human brain function: principles, ... Interpretation Applications