Post on 17-Aug-2020
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!!
bandettini@nih.gov!
“Optimizations” (or at least improvements)!in fMRI !
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
In general !
fMRI before any optimizations (fall of 1991)
2.5 cm !!
TR = 2 sec!TE = 50 ms!One slice!In plane 3.75 x 3.75!
1991! 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�
Methodology
Interpretation Applications
Technology High field strength Coil arrays High resolution Novel functional contrast
Paradigm Designs Processing Methods
Fluctuations / Correlations Dynamics Healthy Brain Organization
Focus of this lecture!
• Field Strength • Echo Time • Spin-echo vs Gradient Echo • Velocity Nulling • RF coil arrays • High Spatial Resolution
• High Temporal Resolution • Choice of Flip Angle • Choice of Slice Thickness • Paradigm Design
• Ultimate Sensitivity? • Separating “good” and “bad” signal in Resting
State fMRI. • Understanding dynamic nonlinarities • Understanding and Using fMRI Patterns
Methodology
Interpretation Applications
Technology High field strength Coil arrays High resolution Novel functional contrast
Paradigm Designs Processing Methods
Fluctuations / Correlations Dynamics Healthy Brain Organization
Focus of this lecture!
Characteristics of the BOLD signal: T2* effect.! !
Contrast at 3T (dR2* = -1.6 1/s)
00.0050.010.0150.020.0250.030.0350.04
0 10 20 30 40 50 60 70 80 90 100
TE (ms)
Co
ntr
ast
60555045403530252015105
Contrast at 1.5T (dR2* = -.8 1/s)
00.0050.010.0150.020.0250.030.0350.04
0 10 20 30 40 50 60 70 80 90 100
TE (ms)
Co
ntr
ast
60555045403530252015105
T2*! T2*!
Contrast depends on:!activation-induced changes in T2* and resting T2*!
Functional Contrast at Optimal TE
00.0050.010.0150.020.0250.030.0350.04
0 10 20 30 40 50 60
T2* (ms)
Con
tras
t
1.5T
3T
Spin-Echo vs. Gradient-Echo
fMRI
Transverse Relaxation!
transverse!magnetization! T2!
T2*!
90°! 180°! 180°!
≈30ms! ≈100ms!
×R2* &×R2
Exchange Regimes
DR2
γ ( ×χ ) Bo[ ][ ]
Fast Intermediate Slow
×R2
×R2*
- 1 >> 1<< 1compartment ! radius <3 µm 3 to 15 µm > 15 µm!
Spin echo vs. Gradient echo!
GE TE = 30 ms
SE TE = 110 ms
×R2* &×R2
Exchange Regimes
DR2
γ ( ×χ ) Bo[ ][ ]
Fast Intermediate Slow
×R2
×R2*
- 1 >> 1<< 1
compartment size 2.5 to 3 µm 3 to 15 µm 15 to ∞ µm
cont
rast
GE
SE
Bolus Injection of Gadolinium
8 µm to 380 µm5 µm to 8 µm
5 µm
Spin-Echo TE = 105 ms
TR = ∞
Gradient-Echo TE = 50 ms
Gradient-Echo functional TE = 50 ms
Spin-Echo functional
TE = 105 ms
3T!
0
0.2
0.4
0.6
0.8
1
1.2
0 20 40 60 80 100 120 140
Intr
avas
cula
rCo
ntrib
utio
n
TE (ms)
1.0
1.5
2.0
2.53.0
3.54.0
0.5
Spin-echo, %HbO2 = 60! !
0
0.2
0.4
0.6
0.8
1
0 20 40 60 80 100 120 140
Intr
avas
cula
rCo
ntrib
utio
n
TE (ms)
1.0
1.5
2.0
2.53.0
3.54.0
0.5
Gradient-echo, %HbO2 = 60!
Source of most contrast in venograms..!
Field strength dependence of intravascular signal !
Bove-Bettis, et al (2004), SMRT
BOLD effect to highlight veins: 3 Tesla!
MRM 30:380-386 (1993)!
Pros and Cons of Spin-Echo
• Increased specificity (esp at high
fields where IV signal is low)
• Less sensitive to rapidly flowing
blood
• Less signal dropout.
• Less slices per TR
• Lower fCNR by x 2 to 4.
• Acquisition window still T2*
• Very large IV signal still present
at most field strengths.
I would only use at 7T if also imaging at high ressolution and interested in something like columns or layers.
…so let’s remove the intravascular signal... Velocity Nulled (or diffusion weighted) fMRI.
8 µm to 380 µm5 µm to 8 µm
5 µm
8 µm to 380 µm5 µm to 8 µm
5 µm
no diffusion weighting! diffusion weighting!
J. L. Boxerman, P. A. Bandettini, K. K. Kwong, J. R. Baker, T. L. Davis, B. R. Rosen, R. M. Weisskoff, The intravascular contribution to fMRI signal change: monte carlo modeling and diffusion - weighted studies in vivo. Magn. Reson. Med. 34, 4-10 (1995).
b = 0!
b = 160!b = 50!
b = 10!
BOLDPerfusionNo
VelocityNulling
VelocityNulling
ASL
GE
SETI(IV) (IV)
Time(sec)
1 2 40 3
Venous inflow!(Perf. No VN)!
Arterial inflow!(BOLD TR < 500 ms)!
Hemodynamic Specificity!
High Field Tradeoffs
• Increased SNR
• Increased functional contrast
• Ability to reduce voxel volume
• Reduced intravascular signal
• At standard resolution, enhanced
sensitivity to fluctuations
• Increased SAR
• Decreased B0 and B1 homogeneity
• (still somewhat prohibitive)
• Increased costs and effort
• At standard resolution, enhanced
sensitivity to fluctuations
Coil Arrays
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.!
Sensitivity vs. Time needed to scan
K. Murphy, J. Bodurka, P. A. Bandettini, How long to scan? The relationship between fMRI temporal signal to noise and the necessary scan duration. NeuroImage, 34, 565-574 (2007) J. Bodurka, F. Ye, N Petridou, K. Murphy, P. A. Bandettini, NeuroImage, 34, 542-549 (2007)
Temporal Signal to Noise Ratio (TSNR) vs. Signal to Noise Ratio (SNR)
3T, birdcage: 2.5 mm3
3T, 16 channel: 1.8 mm3 7T, 16 channel: 1.4 mm3
suggested voxel volume
Going to High Spatial Resolution
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)! Multishot Imaging
T2* decay!
EPI Window 1!
T2* decay!
EPI Window 2!
Excitations 1 2 4 8 Matrix Size 64 x 64 128 x 128 256 x 128 256 x 256
Multi Shot EPI!
30 ms! 10 sec !to !1 min!
Partial k-space imaging
T2* decay!
EPI Window!
A. Jesmanowicz, P. A. Bandettini, J. S. Hyde, Single shot half k-space high resolution EPI for fMRI at 3T. Magn. Reson. Med. 40, 754-762 (1998).
A. Jesmanowicz, P. A. Bandettini, J. S. Hyde, Single shot half k-space high resolution EPI for fMRI at 3T. Magn. Reson. Med. 40, 754-762 (1998). ≈ 5 to 30 ms
Pruessmann, et al. (and Sodickson et al)
SENSE Imaging
3T single-shot SENSE EPI using 16 channels: 1.25x1.25x2mm Cheng, et al. (2001) Neuron,32:359-374
Technology
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
Going to High Temporal Resolution
-101
0 400 800 1200 1600
T2* - Weighted
AV.G
E
Time (sec.)
Signa
l
0
1
0 400 800 1200 1600
T1 - Weighted
Time (sec.)
Signa
l
Flow
BOLD
P. A. Bandettini, K. K. Kwong, T. L. Davis, R. B. H. Tootell, E. C. Wong, P. T. Fox, J. W. Belliveau, R. M. Weisskoff, B. R. Rosen, (1997). “Characterization of cerebral blood oxygenation and flow changes during prolonged brain activation.” Human Brain Mapping 5, 93-109.!
How rapidly can one switch on and off?
P. A. Bandettini,, Functional MRI using the BOLD approach: dynamic characteristics and data analysis methods, in "Diffusion and Perfusion: Magnetic Resonance Imaging" (D. L. Bihan, Ed.), p.351-362, Raven Press, New York, 1995.
Blamire et al.!
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.
50454035302520151050498
502
506
510
514
518
522
.5 sec1 sec2 sec3 sec5 sec
Motor Cortex
Time (sec)
Sign
al
53210.5
Duration (sec)
Bandettini, et al., The functional dynamics of blood oxygenation level contrast in the motor cortex, 12'th Proc. Soc. Magn. Reson. Med., New York, p. 1382. (1993).
How brief of a stimulus can one give?!
5! 10! 15! 20!Time (sec)!
1000 msec!100 msec!34 msec!
R. L. Savoy, et al., Pushing the temporal resolution of fMRI: studies of very brief visual stimuli, onset variability and asynchrony, and stimulus-correlated changes in noise, 3'rd Proc. Soc. Magn. Reson., Nice, p. 450. (1995). !
0
0.5
1
1.5
0 2 4 6 8 10 12 14
250 ms500 ms1000 ms2000 ms
Perc
ent S
igna
l Cha
nge
Time (sec)
0
0.5
1
1.5
0 2 4 6 8 10 12 14
250 ms500 ms1000 ms2000 ms
Perc
ent S
igna
l Cha
nge
Time (sec)
0
0.5
1
1.5
0 2 4 6 8 10 12 14
250 ms500 ms1000 ms2000 ms
Perc
ent S
igna
l Cha
nge
Time (sec)
0
10
20
30
40
50
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Num
ber o
f Vox
els
Delay (sec)
1980
2000
2020
2040
2060
2080
2100
0 2 4 6 8 10 12 14 16 18 20
MRI S
ignal
Time (sec)
P. A. Bandettini, (1999) "Functional MRI" 205-220.
Magnitude
Latency
+ 2 sec
- 2 sec
Venogram
Latency Variation… Hemi-Field Experiment
Right Hemisphere Left Hemisphere
Timing
-2.4"
-1.6"
-0.8"
0"
0.8"
1.6"
2.4"
3.2"
0" 10" 20" 30"
Percent!MR!
Signal!Strength!
Time (seconds)!
Average of 6 runs Standard Deviations Shown!Hemi-field with 500 msec asynchrony"
=!+ 2.5 s!
- 2.5 s!
0 s!
500 ms!500 ms!Right Hemifield!
Left Hemifield!
-!
=!+ 2.5 s!
- 2.5 s!
0 s!
250 ms!250 ms!Right Hemifield!
Left Hemifield!
-!Bottleneck
In Processing (upstream)
Delayed Processing
(downstream)
Hemodynamic Response Modulation
0 5 10 15 20 25 30 0
100
200
300
400
500
Number of runs
8 sec on/off
16 sec on/off Smallest latency Variation Detectable (ms) (p < 0.001)
t
1 run:
-1000 -500 0 500 1000
Num
ber
delay estimate (ms)
�delay = 107ms
11
1% Noise 4% BOLD 256 time pts /run 1 second TR
Even if no hemodynamic variability exists…
11026–11031 PNAS September 26, 2000 vol. 97 no. 20 P. A. Bandettini, (1999) "Functional MRI" 205-220.!
Choosing a Flip Angle Physiological noise effects on the flip angle selection in BOLD fMRI
J. Gonzalez-Castillo a,⁎, V. Roopchansingh b, P.A. Bandettini a,b, J. Bodurka c
a Section on Functional Imaging Methods, Laboratory of Brain and Cognition, National Institute of Mental Health, National Institutes of Health, 10 Center Dr, Bethesda, MD 20892, USAb Functional MRI Facility, National Institute of Mental Health, National Institutes of Health, 10 Center Dr, Bethesda, MD 20892, USAc Laureate Institute for Brain Research, 6655 South Yale Avenue, Tulsa, OK 74136, USA
a b s t r a c ta r t i c l e i n f o
Article history:Received 18 August 2010Revised 29 October 2010Accepted 4 November 2010Available online 10 November 2010
Keywords:Physiological noiseImaging flip angleTemporal signal-to-noise ratioTSNRfMRI
This work addresses the choice of imaging flip angle in blood oxygenation level dependent (BOLD) functionalmagnetic resonance imaging (fMRI). When noise of physiological origin becomes the dominant noise sourcein fMRI timeseries, it causes a nonlinear dependence of the temporal signal-to-noise ratio (TSNR) versussignal-to-noise ratio (SNR) that can be exploited to perform BOLD fMRI at angles well below the Ernst anglewithout any detrimental effect on our ability to detect sites of neuronal activation. We show, bothexperimentally and theoretically, that for situations where available SNR is high and physiological noisedominates over system/thermal noise, although TSNR still reaches it maximum for the Ernst angle, reductionof imaging flip angle well below this angle results in negligible loss in TSNR. Moreover, we provide a way tocompute a suggested imaging flip angle, which constitutes a conservative estimate of the minimum flip anglethat can be used under given experimental SNR and physiological noise levels. For our experimentalconditions, this suggested angle equals 7.63° for the grey matter compartment, while the Ernst angle=77°.Finally, using data from eight subjects with a combined visual-motor task we show that imaging at angles aslow as 9° introduces no significant differences in observed hemodynamic response time-course, contrast-to-noise ratio, voxel-wise effect size or statistical maps of activation as compared to imaging at 75° (an angleclose to the Ernst angle). These results suggest that using low flip angles in BOLD fMRI experimentation toobtain benefits such as (1) reduction of RF power, (2) limitation of apparent T1-related inflow effects,(3) reduction of through-plane motion artifacts, (4) lower levels of physiological noise, and (5) improvedtissue contrast is feasible when physiological noise dominates and SNR is high.
© 2010 Elsevier Inc. All rights reserved.
Introduction
A common practice in gradient recalled-echo (GRE) functionalmagnetic resonance imaging (fMRI) is to select the imaging flipangle to be equal to the Ernst angle (Ernst and Anderson, 1996) forgrey matter. The premise is to select the Ernst angle to maximizethe grey matter signal-to-noise ratio (SNR). Although thisapproach has proven beneficial for spoiled gradient echo anatomicallyoriented applications of MRI, in which physiological noise does notdominate or even contribute significantly (i.e. SNRb50 or in non livingsamples), the same might not be necessarily true for fMRI, in whichnon-thermal noise dominates (Kruger and Glover, 2001; Triantafyllouet al., 2005).
FMRI detects neuronal activity-induced subtle temporal MRIsignal fluctuations that have their origin in the BOLD (BloodOxygenation Level Dependent) phenomenon (Ogawa et al., 1993).These time series contain noise. Specifically, physiological noise ispresent and represents a confounding factor in BOLD fMRI data
(Kruger and Glover, 2001). Depending on the imaging voxel volumeand available SNR, physiological noise frequently becomes thedominant noise source present in fMRI time courses (Bodurkaet al., 2007). The quality of functional MRI data can be characterizedin terms of temporal-SNR (TSNR). This metric is defined andcomputed on a voxel-wise basis as the ratio of the mean steady-state signal of the fMRI time-series to the voxel temporal standarddeviation (Parrish et al., 2000). While SNR is independent ofphysiological noise contributions (i.e., in EPI, an image is collectedwithin 40 ms–faster than most physiologic process–thus minimizingeffects of temporal variations of these), TSNR shows a non-lineardependence on physiological noise contribution. Based on this non-linear dependence of TSNR with physiological noise and the fact thatphysiological noise is MR-signal strength dependent (Kruger andGlover, 2001; Triantafyllou et al., 2005), we hypothesize that thebehavior of SNR and TSNR as a function of imaging flip angle mightdiffer; and that such differences might be exploited to perform fMRIexperiments at imaging angles well below the Ernst angle.
This work studies, both theoretically and experimentally, the TSNRdependence on the flip angle. We provide evidence that, in thepresence of physiologic noise, the TSNR does not follow therelationship defined by the Ernst equation. In fact, we show that
NeuroImage 54 (2011) 2764–2778
⁎ Corresponding author.E-mail address: javier.gonzalez-castillo@nih.gov (J. Gonzalez-Castillo).
1053-8119/$ – see front matter © 2010 Elsevier Inc. All rights reserved.doi:10.1016/j.neuroimage.2010.11.020
Contents lists available at ScienceDirect
NeuroImage
j ourna l homepage: www.e lsev ie r.com/ locate /yn img
How do we typically select flip angle? Ernst Angle…
Θ = Cos-1 (e -TR/T1)
Fig. 2. (A) TSNR vs. Flip angle for T1=T1,GM=1.34 s, SNRo=SNRo,GM=652 and different values of ! ranging from 0 to 0.05. For each curve, a square marks the angle below the Ernstangle for which TSNR has decreased to 50% from its maximum value ("50%). (B) Evolution of "50% with !. The higher the amount of physiological noise present, the flatter the TSNRcurve and consequently the lower the angle for which TSNR reaches 50% of its maximum value.
Fig. 3. Simulations of the suggested fMRI flip angle ("S). The black dotted line shows a linear behavior for the Silicone Oil Phantom. The red and blue continuous lines show behaviorfor grey and white matter respectively. Ernst angle for these two tissue compartments is marked with filled dots, while other exemplary angles are masked as transparent dots. Thesuggested flip angle for both tissue compartments are also marked in each curve as filled squares with black outline.
2767J. Gonzalez-Castillo et al. / NeuroImage 54 (2011) 2764–2778
better contrast at lower flip angles translates in easier segregation oftissue compartments.
Discussion
Physiological noise is a major source of undesired variance inBOLD fMRI time courses in a vast majority of experimental situations(Kruger and Glover, 2001; Kruger et al., 2001; Triantafyllou et al.,2005; Bodurka et al., 2007). We have investigated, both theoreticallyand experimentally, the effect that MR-signal strength-dependentphysiological noise exerts on BOLD fMRI temporal signal to noise
ratio (TSNR) as a function of the flip angle in situations wherephysiological noise constitutes a dominant source of time coursevariance. We have scanned 8 subjects at a commonly used BOLDfMRI voxel volume of 3.75!3.75!4 mm3, where physiological noiseis the dominant source of time course variance (Bodurka et al.,2007); and physiological noise introduces a non-linear dependencein TSNR, which translates into a flattening of the TSNR vs. flip anglecurve. We have also demonstrated that this TSNR behavior can beexploited to perform BOLD-fMRI at flip angles other than the Ernstangle with no detrimental effects in our ability to detect statisticallysignificant neuronal activations.
Fig. 7. Averaged hemodynamic response across all eight subjects for all flip angles in three different anatomically defined ROIs: right visual cortex, left visual cortex and left primarymotor cortex. The top panel shows 3D renderings of the ROIs. Themiddle panel shows estimations of the hemodynamic response without intensity normalization (i.e., only constant,linear and quadratic trends were removed). The bottom panel shows estimations of hemodynamic response in terms of signal percent change. These were obtained by means ofintensity normalization prior to the detrending step.
2772 J. Gonzalez-Castillo et al. / NeuroImage 54 (2011) 2764–2778
and/or elapsed time between scans; they provide support for theclaim that using angles other than the Ernst angle do not cause anydisadvantage when attempting to detect BOLD related activations, atleast for the task and the range of angles under consideration in thisstudy.
Evaluation of plausible flip angle effects on the estimation ofregression coefficients (!) at single-subject statistical analysis alsorendered not significant. This result suggests that second levelstatistical analysis (group analysis), which uses as input ! coefficientsfrom individual subjects, should not be negatively affected by the useof angles other than the Ernst angle. Scatter plot and correlationanalysis results show that a voxel-wise linear relation exists between! coefficients at "=75°(closest available angle to the Ernst angle forGM at 3 T and TR=2 s) and all other angles. Moreover, we showedthat the slope (S) and constant term (C) of this linear relationship arenot significantly different from the ideal case (S=1, C=0) in whichvoxel-wise ! coefficients would be identical across flip angles.Although deviations from the ideal case exists, the fact that theseare not significant or consistent across ROIs, suggest that thesevariations are solely the result of random inter-scan variations notaccounted for by the regression analysis; and not a systematic effectattributable to changes in the flip angle.
Although the analysis on how imaging flip angle affects hemody-namic response, CNR, activation extent and ! coefficients wasperformed on data collected using a specific experimental sensorimotorepoch design, we believe that the main conclusion of such analysis–namely that flip angle has no systematic effect on the ability to detectBOLD-based neuronal activity–remains valid for other types of fMRIexperimental paradigms such as event-related designs, high-ordercognitive tasks, etc. We construct this claim on the base that changes to
the experimental task are not expected tomodify the TSNR vs.flip anglenon-linear relationship described in this work. Nonetheless, prior toadoption of a low imaging angle for a set of experiments, researchersare encouraged to confirm that physiological noise is the dominantsource of noise in their setting (#p/#oNN1) and that SNR levels aresufficiently high. One way to obtain such confirmation is to computethe suggested imaging flip angle using Eq. (10) (or its non-approximated version Eq. (9)). Inputs to these formulas includeexperimental measures of physiological noise and SNR levels specificto the experimental settings under consideration. If both requirementsare satisfied, the suggested flip angle will be lower than the Ernst angle.On the other hand, if one or both of these requirements are not satisfiedthe computed suggested angle will be equal or greater than the Ernstangle (see Fig. 4 for a case when physiological noise decreases to levelsin which it is no longer the dominant noise source) and researchersshould consider using the Ernst angle. As a general guideline,experiments conducted at field strengths equal or greater than 3Tesla, using receiver array-coils of eight or more elements and voxelssize in the vicinity of 50 mm3 (3.75!3.75!4.00 mm) will lead tosuggested imaging flip angles well below the Ernst angle.
We believe these results have important implications for exper-imental fMRI, as the use of small flip angles provides importantadditional benefits such as better tissue contrast, less inflow effects(Gao et al., 1996), less through-plane motion artifacts, lowerphysiological noise levels, shorter scanning times, and reduced levelsof radio-frequency (RF) energy deposition. All these benefits arisethrough the following mechanisms. The flip angle for an MRexperiment is computed by:
θ = γ ! RF pulse amplitude ! RF pulse duration
Fig. 12. (A) Experimental and Theoretical curves describing dependence of tissue contrast with flip angle for three tissue contrast of interest: GM vs. WM (ΔSGM,WM), GM vs. CSF(ΔSGM,CSF), and WM vs. CSF (ΔSWM,CSF). (B) Axial slices, after steady-state has been reached, for all acquired flip angles for an exemplary subject.
2776 J. Gonzalez-Castillo et al. / NeuroImage 54 (2011) 2764–2778
Effect of Slice Thickness on TSNR
P. S. F. Bellgowan, P. A. Bandettini, P. van Gelderen, A. Martin, J. Bodurka, Improved BOLD detection in the medial temporal region using parallel imaging and voxel volume reduction. NeuroImage, 29, 1244-1251 (2006)
Methodology
Interpretation Applications
Technology High field strength Coil arrays High resolution Novel functional contrast
Paradigm Designs Processing Methods
Fluctuations / Correlations Dynamics Healthy Brain Organization
Focus of this lecture!
Neuronal Activation Input Strategies
1. Block Design
2. Frequency Encoding
3. Phase Encoding
4. Event-Related
5. Orthogonal Block Design
6. Free Behavior Design.
Neuronal Activation Input Strategies
1. Block Design
2. Frequency Encoding
3. Phase Encoding
4. Event-Related
5. Orthogonal Block Design
6. Free Behavior Design.
P. A. Bandettini, A. Jesmanowicz, E. C. Wong, J. S. Hyde, Processing strategies for time-course data sets in functional MRI of the human brain. Magn. Reson. Med. 30, 161-173 (1993).
Neuronal Activation Input Strategies
1. Block Design
2. Frequency Encoding
3. Phase Encoding
4. Event-Related
5. Orthogonal Block Design
6. Free Behavior Design.
Tapping left and right fingers at two different “on/off” frequencies
P. A. Bandettini, A. Jesmanowicz, E. C. Wong, J. S. Hyde, Processing strategies for time-course data sets in functional MRI of the human brain. Magn. Reson. Med. 30, 161-173 (1993).
Neuronal Activation Input Strategies
1. Block Design
2. Frequency Encoding
3. Phase Encoding
4. Event-Related
5. Orthogonal Block Design
6. Free Behavior Design.
E. A. DeYoe, P. A. Bandettini, J. Nietz, D. Miller, P. Winas, Functional magnetic resonance imaging (FMRI) of the human brain. J. Neuroscience Methods 54, 171-187 (1994).
E. A. DeYoe, G. Carman, P. Bandettini, G. S., W. J., R. Cox, D. Miller, J. Neitz, Mapping striate and extrastriate visual areas in human cerebral cortex. Proc. Nat'l. Acad. Sci. 93, 2282-2386 (1996).
E. A. DeYoe, G. Carman, P. Bandettini, G. S., W. J., R. Cox, D. Miller, J. Neitz, Mapping striate and extrastriate visual areas in human cerebral cortex. Proc. Nat'l. Acad. Sci. 93, 2282-2386 (1996).
Neuronal Activation Input Strategies
1. Block Design
2. Frequency Encoding
3. Phase Encoding
4. Event-Related
5. Orthogonal Block Design
6. Free Behavior Design.
R. L. Buckner, P. A. Bandettini, K. M. O'Craven, R. L. Savoy, S. E. Peterson, M. E. Raichle, T. L. Brady, B. R. Rosen, fMRI detection and time course of distributed cortical activations during single trials of a cognitive task. Proc. Nat'l. Acad. Sci. USA 93, 14878-14883 (1996).
P. A. Bandettini, R. W. Cox. Functional contrast in constant interstimulus interval event - related fMRI: theory and experiment. Magn. Reson. Med. 43: 540-548 (2000). !
20, 20! 12, 2! 10, 2! 8, 2! 6, 2! 4, 2! 2, 2!
( ISI, SD )!
S1!
S2!
Contrast to Noise Images!
P. A. Bandettini, R. W. Cox. Functional contrast in constant interstimulus interval event - related fMRI: theory and experiment. Magn. Reson. Med. 43: 540-548 (2000). !
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!
Detectability vs. Average ISI
0 5 10 15 20 25 30 35 40
average ISI (s)
Det
ecta
bilit
y
SD = 1000 ms.
SD = 250 ms.
SD = 4000 s.
R. M. Birn, R. W. Cox, P. A. Bandettini, Detection versus estimation in Event-Related fMRI: choosing the optimal stimulus timing. NeuroImage 15: 262-264, (2002).
Estimation accuracy vs. average ISI
0 5 10 15 20 25 30 35 40 0
5
10
15
20
average ISI (sec)
Est
imat
ion
Acc
urac
y
SD = 4000 ms.
SD = 1000 ms.
SD = 250 ms.
R. M. Birn, R. W. Cox, P. A. Bandettini, Detection versus estimation in Event-Related fMRI: choosing the optimal stimulus timing. NeuroImage 15: 262-264, (2002).
fMRI during tasks that involve brief motion!
motion! BOLD response!
task!
BOLD response!
t!
motion!
task!
Blocked Design!
Event-Related Design!
R. M. Birn, P. A. Bandettini, R. W. Cox, R. Shaker, Event - related fMRI of tasks involving brief motion. Human Brain Mapping 7: 106-114 (1999). !
Speaking - Blocked Trial!
Expected!Response!
motion!
BOLD!response!
t!
t!
R. M. Birn, P. A. Bandettini, R. W. Cox, R. Shaker, Event - related fMRI of tasks involving brief motion. Human Brain Mapping 7: 106-114 (1999). !
Speaking - ER-fMRI!
avg!
avg!
Expected!Response!
R. M. Birn, P. A. Bandettini, R. W. Cox, R. Shaker, Event - related fMRI of tasks involving brief motion. Human Brain Mapping 7: 106-114 (1999). !
0! 10! 20! 30! 40! 50! 60! 70! 80! 90! 100!20!
10!
0!
10!
20!
Detection of BOLD (t-stat)!
Det
ectio
n of
Mot
ion
(t-s
tat)!
Blocked!
Event-Related!constant ISI!(SD=1s, ISI=15s)!
Event-Related!varying ISI!
min SD=1s! min SD=3s!
min SD=5s!
min SD=7s!
min SD=9s!SD=2s!
SD=4s!
SD=6s!
SD=8s!
Overt Responses - Simulations!
R.M. Birn, R. W. Cox, P. A. Bandettini, NeuroImage, 23, 1046-1058 (2004) !
SD = stimulus duration!
Mo
re M
oti
on
Art
ifac
ts!
Better BOLD Detection!
Blocked design (30s/30s)!
Blocked design (10s/10s)!
Event-related!Varying ISI (1s min. SD)!
Event-related!Varying ISI (5s min. SD)!
Event-related!constant ISI (1s. SD, 15sISI)!
t-stat.!
0!
-15!
15!
Overt Responses!
Neuronal Activation Input Strategies
1. Block Design
2. Frequency Encoding
3. Phase Encoding
4. Event-Related
5. Orthogonal Block Design
6. Free Behavior Design.
Example of a Set of Orthogonal Contrasts for Multiple Regression!
NonselectiveVisualStimulation
CTL DELAY
FACEWM
DELAYI T I I T I
HOUSEWM
DELAY
Faces & Housesvs Ctl. Stimuli
Ctl. Stim. vs.Ctl. Response
Encoding vs. Recognition
Anticipatorydelays vs ITIs
I T I
Memory delays vs. ctl. delays
Face Stimuli vsHouse Stimuli
Face WM delays vsHouse WM delays
Courtney et al.
Neuronal Activation Input Strategies
1. Block Design
2. Frequency Encoding
3. Phase Encoding
4. Event-Related
5. Orthogonal Block Design
6. Free Behavior Design.
What is the Ultimate Sensivitity of fMRI?
Is#the#whole#brain#ac0vated#by#even#simple#tasks?#
#• 9"hours"of"averaging"• Unconstrained"response"model"• Clustering"(K9means"or"Hierarchical)##
J. Gonzalez-Castillo, Z. Saad, D. A. Handwerker, P. A. Bandettini, Whole-brain, time-locked activation with simple tasks revealed using massive averaging and model-free analysis. Proceedings of the National Academy of Sciences 109, 14: pp. 5487-5492 (2012)
The#univariate#approach#
SUBJECT#EXPERIMENTAL#PARADIGM##
PREDICTED#HEMODYNAMIC#
RESPONSE#
=#X#
HRF#
STATISTICAL#MAP#OF#
ACTIVATION#
DATA#PREO#
PROCESSING#
RESPONSE#
MODEL#
STATISTICAL#
ANALYSIS#
…"
Data#averaging#
EXPERIMENTAL#PARADIGM## SUBJECT#
…"
X#
PREDICTED#HEMODYNAMIC#
RESPONSE#
=#HRF#
STATISTICAL#MAP#OF#
ACTIVATION#
DATA#PREO#
PROCESSING#
RESPONSE#
MODEL#
STATISTICAL#
ANALYSIS#
…"
…"
…"
Mul0ple#Response#Models#
EXPERIMENTAL#PARADIGM## SUBJECT#
…"
X#
PREDICTED#HEM.#RESPONSES#
=#HRF#
STATISTICAL#MAP#OF#
ACTIVATION#
PREOPROCESSING#
RESPONSE#
#MODELS#
STATISTICAL#ANALYSIS#
…"
…"
…"
SUSTAINED#RESPONSE#MODEL#
BLOCK#DESIGN#&#HEMIFIELD#VISUAL#STIMULATION#
ONSET/OFFSET#RESPONSE#MODEL#
Predic0ve#Response#Model#effect#on#fMRI#Results#(III)#
Uludag"et"al."Magn"Reson"Imaging."2008"Sep;26(7):863@9.""
DIFFERENT#RESPONSE#SHAPES#ARE#PRESENT#ACROSS#DIFFERENT#
REGIONS#OF#THE#BRAIN#FOR#A#SINGLE#STIMULUS#TYPE#25"""30"""35"""40"""45"""50"""55""…"
+"
Navg"
IS#THE#SPARSENESS#OF#FMRI#ACTIVATIONS#REAL?#
OR#
IS#IT#THE#RESULT#OF#INSUFFICIENT#TSNR#+#OVERLY#STRICT#RESPONSE#MODELS?#
Saad"et"al"
3"Healthy"Volunteers:"1M/2F;"Age"="27"±"2.5" 3T"GE"Signa"HDx" Anatomical"Scan:"MPRAGE"|".9x.9x1.2"mm3"|"192"Slices" FuncZonal"Scans:"GRE@EPI"
• TR/TE"="2s/30ms"• In@Plane"Res"="64x64""• #Slices"="32"Oblique"
• "FOV"="240mm"• "Slice"Thickness"="3.8"mm""• "Flip"Angle"="75°"
Experimental#Methods#(I)#
3x" ANATOMICAL"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
"""""(1x""""""""""")"
1"
VISIT"
2"
3"
4"
5"
6"
7"
8"
9"
10"
:"
:"
:"
:"
:"
:"
:"
:"
:"
:"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
+""(10x"""""""""""""""""""""""""""""")"
FUNCTIONAL"SCANS"
100"FUNCTIONAL"RUNS/SUBJECT"
500"TRIALS/SUBJECT"
9"HOURS"OF"DATA/SUBJECT"
Experimental#Methods#(II)#
X"100""(QA"Axial"EPIs)"
Data#Analysis#
DATA#PREOPROCESSING#
Remove"Physiological"Noise"
Head"MoZon"CorrecZon"
Slice"Timing"CorrecZon"
Inter@run"Co@registraZon"
Discard"IniZal"5"Volumes"
Remove"MoZon"&"1st"Der."
Intensity"NormalizaZon"
DATA#AVERAGING#
Navg"="1"<"@@">"Navg"="100"
10"Random"PermutaZons"per"Navg"Level"
Example"Navg"="5"
Data#Analysis#
DATA#PREOPROCESSING#
Remove"Physiological"Noise"
Head"MoZon"CorrecZon"
Slice"Timing"CorrecZon"
Inter@run"CoregistraZon"
Discard"IniZal"5"Volumes"
Remove"MoZon"&"1st"Der."
Intensity"NormalizaZon"
DATA#AVERAGING#
Navg"="1"<"@@">"Navg"="100"
10"Random"PermutaZons"per"Navg"Level"
Example"Navg"="5"
STATISTICAL#ANALYSIS#
SUSTAINED#RESPONSE#
ONLY#(SUS)#
ONSET#+#SUSTAINED#+#
OFFSET#RESPONSE#(SUS)#
UNCONSTRAINED#MODEL#
(UNC)#
CLUSTERING#ANALYSIS#
Results:#TSNR#vs.###Averaged#Scans#
TSNRWM"="339"""1"
TSNRWM"="2218"""100"
TSNRWM#INCREASED#BY#APPROX.#
A#FACTOR#OF#6#FROM##
Navg#=#1#TO#Navg#=#100#
Number"of"Averaged"Scans"(Navg)"
TSNR"
Results:#TimeOseries#in#Primary#Visual#Cortex#
INDIVIDUAL#RUNS#
100"
98"
102"
104"
LEFT" RIGHT" LEFT"RIGHT"
AVERAGING#
Time"(s)""""30"""50"""""""""90"""110"""""150"170""""""210"230""""""270"290""""""""
Time"(s)"""""""30"50"""""""""90""110""""""150"170""""""210"230"""""270"290""""""""
Rest" Rest" Rest" Rest" Rest" Rest"TASK# TASK# TASK# TASK# TASK#
0s" 30s" 50s" 90s" 110s" 150s" 170s" 210s" 230s" 270s" 290s" 340s"
Gonzalez@CasZllo"J,"Saad"ZS,Handwerker"DA,"InaZ""SJ,"Brenowitz"N,"Bandepni"PA,"PNAS"(in"press)""
Results:#TimeOseries#in#Anterior#Insular#Cortex#
INDIVIDUAL#RUNS#
100"
99"
101"
LEFT" RIGHT" LEFT"RIGHT"
AVERAGING#
Time"(s)""""30"""50"""""""""90"""110"""""""150"170"""""""210"230"""""""270"290""""""""
Time"(s)""""30"""50"""""""""90"""110"""""""150"170"""""""210"230"""""""270"290""""""""
Rest" Rest" Rest" Rest" Rest" Rest"TASK# TASK# TASK# TASK# TASK#
0s" 30s" 50s" 90s" 110s" 150s" 170s" 210s" 230s" 270s" 290s" 340s"
Gonzalez@CasZllo"J,"Saad"ZS,Handwerker"DA,"InaZ""SJ,"Brenowitz"N,"Bandepni"PA,"PNAS"(in"press)""
Results:#TimeOseries#in#Primary#Auditory#Cortex#
INDIVIDUAL#RUNS#
100"
99.5"
100.5"
LEFT" RIGHT" LEFT"RIGHT"
AVERAGING#
Time"(s)"
Rest" Rest" Rest" Rest" Rest" Rest"TASK# TASK# TASK# TASK# TASK#
0s" 30s" 50s" 90s" 110s" 150s" 170s" 210s" 230s" 270s" 290s" 340s"
"""30"""50"""""""""90"""110"""""""150"170"""""""210"230"""""""270"290""""""""
Time"(s)""""30"""50"""""""""90"""110"""""""150"170"""""""210"230"""""""270"290""""""""
Gonzalez@CasZllo"J,"Saad"ZS,Handwerker"DA,"InaZ""SJ,"Brenowitz"N,"Bandepni"PA,"PNAS"(in"press)""
Results:#TimeOseries#in#ParietoOOccipital#Junc0on#
INDIVIDUAL#RUNS#
100"
98"
102"
LEFT" RIGHT" LEFT"RIGHT"
AVERAGING#
Time"(s)"
Rest" Rest" Rest" Rest" Rest" Rest"TASK# TASK# TASK# TASK# TASK#
0s" 30s" 50s" 90s" 110s" 150s" 170s" 210s" 230s" 270s" 290s" 340s"
"""30"""50"""""""""90"""110"""""""150"170"""""""210"230"""""""270"290""""""""
Time"(s)""""30"""50"""""""""90"""110"""""""150"170"""""""210"230"""""""270"290""""""""
Gonzalez@CasZllo"J,"Saad"ZS,Handwerker"DA,"InaZ""SJ,"Brenowitz"N,"Bandepni"PA,"PNAS"(in"press)""
Responses#0meOlocked#with#the#task#were#observed#in#over#90%#of#the#
voxels#for#all#three#subjects#
Results:#BOLD#responses#are#present#all#over#the#brain#
Area#of#Ac0va0on#vs.##Scans#
pBonf<0.05"
SUSTAINED#M
ODEL#
Subject"1"
T"
@100"
100"
Gonzalez@CasZllo"J,"Saad"ZS,Handwerker"DA,"InaZ""SJ,"Brenowitz"N,"Bandepni"PA,"PNAS"(in"press)""
WITHINOSUBJECT#AVERAGED#
RESPONSES#ACROSS#ALL#
RUNS#AND#TRIALS#
ARE#RESPONSE#SHAPES#RANDOMLY#DISTRIBUTED#ACROSS#THE#BRAIN?##
OR#
##
DO#THEY#CLUSTER#IN#A#FUNCTIONALLY/ANTOMICALLY#MEANINGFUL#
MANNER?#
CLUSTERING#
Clustering#Analysis:#KOmeans#applied#to#fMRI#data#
Time"series""of"length"30"for"each"of"N"voxels"(e.g.,"all"GM"voxels)" Pearson"CorrelaZon"Distance""D"="1"–"r""
D"="0"(r"="1)"if"Zme"series"from"2"voxels"are"perfectly"correlated" D"="2"(r"="@1)"if"Zme"series"from"2"voxels"are"perfectly"anZ@correlated"
K"="Set"a"priori"by"the"experimenter"
INPUT#
KOMEANS#
K=2"
The"output"consists"on"K"clusters,"each"defined"by:" Set"of"Voxels"(not"necessarily"conZguous)" Centroid"Time"Series"="Average"of"Zme"series"across"all"voxels"in"the"cluster"O
UTPUT#
NO#SPATIAL#INFORMATION#ENTERS#THE#CLUSTERING#ALGORITHM#
Clustering#Analysis:#Whole#Brain#GM#Results#
NOT#RANDOMLY#DISTRIBUTED#IN#SPACE#
SYMETRICAL#ACROSS#HEMISPHERES#
FUNCTIONALLY#&#ANATOMICALLY#MEANINGFUL#
REPRODUCIBLE#PARCELLATIONS#ACROSS#SUBJECTS#
SUBJECT#03#–#K=20#
Clustering#Analysis:#Whole#Brain#GM#Results#
NOT#RANDOMLY#DISTRIBUTED#IN#SPACE#
SYMETRICAL#ACROSS#HEMISPHERES#
FUNCTIONALLY#&#ANATOMICALLY#MEANINGFUL#
REPRODUCIBLE#PARCELLATIONS#ACROSS#SUBJECTS#
SUBJECT#03#–#K=20#
Clustering#Analysis:#Whole#Brain#GM#Results#
NOT#RANDOMLY#DISTRIBUTED#IN#SPACE#
SYMETRICAL#ACROSS#HEMISPHERES#
FUNCTIONALLY#&#ANATOMICALLY#MEANINGFUL#
REPRODUCIBLE#PARCELLATIONS#ACROSS#SUBJECTS#
SUBJECT#03#–#K=20#
Clustering#Analysis:#Whole#Brain#GM#Results#
NOT#RANDOMLY#DISTRIBUTED#IN#SPACE#
SYMETRICAL#ACROSS#HEMISPHERES#
FUNCTIONALLY#&#ANATOMICALLY#MEANINGFUL#
REPRODUCIBLE#PARCELLATIONS#ACROSS#SUBJECTS#
SUBJECT#03#–#K=20#
Clustering#Analysis:#Whole#Brain#GM#Results#
NOT#RANDOMLY#DISTRIBUTED#IN#SPACE#
SYMETRICAL#ACROSS#HEMISPHERES#
FUNCTIONALLY#&#ANATOMICALLY#MEANINGFUL#
REPRODUCIBLE#PARCELLATIONS#ACROSS#SUBJECTS#
SUBJECT#03#–#K=20#
Clusters#as#a#func0on#of#K#
K=02"
K=05"
K=10"
K=15"
K=20"
K=25"
K=30"
K=70"
SUBJECT"01"
Clusters#as#a#func0on#of#Clustering#Algorithm#(I)#
KOMEANS#(d=Correla0on)#–#K#=#05#
HIERARCHICAL#CLUSTERING#
#(link=Ward,#d=Euclidean)#–#K#=#05#
CPCC#=#0.84#
Clusters#as#a#func0on#of#Clustering#Algorithm#(II)#
KOMEANS#(d=Correla0on)#–#K#=#20#
HIERARCHICAL#CLUSTERING#
#(link=Ward,#d=Euclidean)#–#K#=#20#
CPCC#=#0.84#
Clustering#Analysis:#Subcor0cal#GM#Results#
FREESURFER#ANATOMICAL#SEGMENTATION#
Thalamus" "Putamen" ""Caudate" Pallidus" N."Accumbens"
(C)"
KOMEANS#CLUSTERING#(K=5)#
Gonzalez@CasZllo"J,"Saad"ZS,Handwerker"DA,"InaZ""SJ,"Brenowitz"N,"Bandepni"PA,"PNAS"(in"press)""
Future#Direc0ons#(I)#
K#=#02#(1)#Evaluate#the#Stability#of#the#Clusters#
SET#02#
KOMEANS#(Correla0on)#–#K#=#15#
SET#01#
KOMEANS#(Correla0on)#–#K#=#15#
Gonzalez@CasZllo"J,"Saad"ZS,Handwerker"DA,"InaZ""SJ,"Brenowitz"N,"Bandepni"PA,"PNAS"(in"press)""
Future#Direc0ons#(II)#
K#=#02#(2)#Evaluate#how#these#clusters#relate#to#Res0ng#State#Networks#
RESTING#STATE#
NETWORKS#
TASKOBASED#
NETWORKS#
KOMEANS#(Correla0on)#–#K#=#15#
Smith"et"al."Proc"Natl"Acad"Sci"U"S"A."2009"Aug"4;106(31):13040@5"
Gonzalez@CasZllo"J,"Saad"ZS,Handwerker"DA,"InaZ""SJ,"Brenowitz"N,"Bandepni"PA,"PNAS"(in"press)""
Separating “good” and “bad” signal !
in Resting State fMRI.!Res0ng#state#clustering#in#mul0ple#and#single#subjects.#
• Mul=9echo"denoising"• Hierarchical"clustering##
P. Kundu, S. J. Inati, J. W. Evans, W.-M. Luh, P. A. Bandettini, Differentiating BOLD and non-BOLD signals in fMRI time series using multi-echo EPI. NeuroImage 60, pp. 1759-1770 (2012)
Rest: seed voxel in motor cortex
Activation: correlation with reference function
B. Biswal et al., MRM, 34:537 (1995)
Resting State Correlations
142
("resting state" OR fluctuations OR "spontaneous oscillations" OR "endogenous oscillations") AND fMRI
Endogenous Oscillations
143
De Luca, Neuroimage 29(4) 2006
Visual
Visuospatial Executive
Sensory Auditory
Dorsal Pathway
Ventral Pathway
Sources of time series fluctuations • Blood, brain and CSF pulsation
• Vasomotion
• Breathing cycle (B0 shifts with lung expansion)
• Bulk motion
• Scanner instabilities
• Changes in blood CO2 (changes in breathing)
• Spontaneous neuronal activity
Noise and Fluctuations
Fig. 4. Pie charts showing the fMRI data variance explained (VE, %, upper bold) by nonthermal noise sources 1–4, thermal noise and spontaneous activity. We also show fMRI signal change (SC, %, lower italic) attributed to the same noise sources. Average (S.E.) values across subjects are shown. The contribution of thermal noise at the ROI level was negligible.
Bianciardi et al. Magnetic Resonance Imaging 27: 1019-1029, 2009
Toes Fingers Thumb Eyebrow Tongue
Sources of time series fluctuations: • Blood, brain and CSF pulsation
• Vasomotion
• Breathing cycle (B0 shifts with lung expansion)
• Bulk motion
• Scanner instabilities
• Changes in blood CO2 (changes in breathing)
• Spontaneous neuronal activity
148
Breath-holding
0
5
Z
Cue 0 50 100 150 200 250 300 350
5 %
Respiration
MR Signal
Group Maps (N = 7)
Anatomy Breath-hold response (average Z-score)
time (s)
R.M. Birn, J. A. Diamond, M. A. Smith, P. A. Bandettini, NeuroImage, 31, 1536-1548
Spontaneous changes in respiration and end-tidal CO2
Respiration
Respiration Volume / Time (RVT)
0 50 100 150 200 250 300 350 time (s)
RVT = max - min T
T max
min 300 310 315
time (s) 305
time (s) 50 100 150 200 250 300 350
0 50 100 150 200 250 0
2 CO2
RVT
0 10 -0.5 0
0.5 1
Shift (s)
CC
-20 -10 RVT precedes end tidal CO2 by 5 sec. 150
Respiration induced signal changes Rest Breath-holding
(N=7) 0
5
Z
0
-4
Z
0 100 200 300 Time (sec)
BOLD
Respiration
RVT
0 100 200 300 Time (sec)
BOLD
Respiration
Cue
R.M. Birn, J. A. Diamond, M. A. Smith, P. A. Bandettini, NeuroImage, 31, 1536-1548 (2006) 151
RVT Correlation Maps & Functional Connectivity Maps
Group (n=10)
Resting state correlation with signal from posterior cingulate
-10
10
Z
0
6
|Z|
Resting state correlation with RVT signal
R.M. Birn, J. A. Diamond, M. A. Smith, P. A. Bandettini, NeuroImage, 31, 1536-1548 (2006)
-10
10
Z
-10
10
Z
Constant Respiration Rate
Group (n=10)
Effect of Respiration Rate Consistency on Resting Correlation Maps
Spontaneously Varying Respiration Rate
R.M. Birn, J. A. Diamond, M. A. Smith, P. A. Bandettini, NeuroImage, 31, 1536-1548 (2006)
Respiration Changes vs. BOLD
time (s) 350 0 50 100 150 200 250 300
RVT
fMRI Signal
0 50 100 150 200 250 300 350 time (s)
How are the BOLD changes related to respiration variations?
?
154
fMRI response to a single Deep Breath
Respiration
40s
0
1
0 20 40 60 time (s)
� S
igna
l (%
)
-1
fMRI Signal
Deep Breath
0 10 20 30 40 time (s)
RRF(t) = 0.6 t 2.1 e
t 1.6 0.0023 t 3.54
e t
4.25
…
Respiration Response Function
(RRF)
deconv.
R.M. Birn, M. A. Smith, T. B. Jones, P. A. Bandettini, NeuroImage, 40, 644-654. (2008) 155
Respiration response function predicts BOLD signal associated with breathing changes better than activation response function.
Breath-holding
time (s)
Sign
al (%
)
20s 40-60s 0 100 200 300 -2
0
2
4
Rate Changes
20s 40s
time (s) Si
gnal
(%)
0
-3 0 100 200 300
Depth Changes
20s 40s
time (s)
Sign
al (%
)
-4
0
3
0 100 200 300
R.M. Birn, M. A. Smith, T. B. Jones, P. A. Bandettini, NeuroImage, 40, 644-654. (2008) 156
BOLD magnitude calibration Before
Calibration After
Calibration Respiration-induced �S
Breath Hold
Rest
Depth Change
Rate Change
BOLDcalib = %�S (BOLD) %�S (Resp)
TE#Dependence*of*BOLD*
• BOLD*T2**signal*has*echo#time*(TE)*dependence,*such*that*the*percent*signal*change*of*a*BOLD*signal*time*course*scales*linearly*with*TE*1*
• Acquiring*multi#echo*(ME)*fMRI*enables*analysis*of*TE#dependence*for*any*signal,*task#correlated*or*spontaneous*2*
• TE#dependence*can*be*quantified,*per#voxel,*as*an*F#statistic*for*the*TE#dependence*model*
*1 Menon,"R.,"Ogawa,"S.,"Tank,"D.,"Ugurbil,"K.,"1993."Tesla"gradient"recalled"echo"characterisZcs"of"phoZc"sZmulaZon@induced"signal"changes"in"the"human"primary"visual"cortex."MagneZc"Resonance"in"Medicine"30,"380"*2 PelZer,"S.J.,"Noll,"D.C.,"2000."Analysis"of"fMRI"signal"and"noise"component"TE"dependence."Neuroimage"11,"S623."*
�Si
Si= ��R⇤
2TEi
TE-Dependence model
TE@Dependence"
κ Rank
κ
High κ
Low κ
Regress"out"all""low"ĸ"components"to"de@noise""
Effect"of"de@noising"on"funcional"connecZvity"measures"Denoising"
Kundu, Guillod, Inati, Luh, Bandettini
Detailed whole brain functional organization from resting state fluctuations?
Rest%1% Rest%2%(color%matched%to%Rest%1)%
Test@retest"of"individual#clustering"at"350"clusters"
Motor" Motor"Angular"Gyrus"
Angular"Gyrus"
Test@retest"of"group"clustering"at"350"clusters"
Rest%1% Rest%2%(color%matched%to%Rest%1)%
Two other issues with imaging resting state fluctuations: 1. Global signal correction or not?
2. Short range correlations may be scanner-related.
The issue of global signal regression
K. Murphy, R. M. Birn, D. A. Handwerker, T. B. Jones, P. A. Bandettini, NeuroImage, 44, 893-905 (2009)
The issue of correlation across voxels due scanner instabilities
N. Kriegeskorte, J. Bodurka, P. Bandettini, International Journal of Imaging Systems and Technology, 18 (5-6), 345-349 (2008)
Methodology
Interpretation Applications
Technology High field strength Coil arrays High resolution Novel functional contrast
Paradigm Designs Processing Methods
Fluctuations / Correlations Dynamics Healthy Brain Organization
Focus of this lecture!
Neuronal Activation
Hemodynamics ? ? ?
Measured Signal
Noise
?
Interpretation
Understanding Dynamic Nonlinearities!
In fMRI!
Nonlinearity of BOLD response!
250 ms! 500 ms! 1000 ms! 2000 ms!
500 ms! 1000 ms! 2000 ms! 4000 ms!
measured! ideal (linear)!
visual!stimulation!
motor!task!
Logothetis et al. Nature, 412, 150-157!
Bandettini and Ungerleider, Nature Neuroscience, 4, 864-866!
Duty Cycle Effects!
R.M. Birn, P. A. Bandettini, The effect of stimulus duty cycle and "off" duration on BOLD response linearity. NeuroImage, 27, 70-82 (2005)
Understanding and Using !
Activation Patterns!
In fMRI!
Ventral temporal category representations
Object categories are associated with distributed representations in ventral temporal cortex
Haxby et al. Nature 2001
response patterns
stimuli ...
...
ROI in Brain
dissimilarity matrix
0 0 0 0 0 0 0 0 0 0 0 0
96
96
compute dissimilarity (1-correlation across space)
96
Dissimilarity Matrix Creation
N. Kriegeskorte, et al, Neuron 60, 1-16 (2008)
Visual Stimuli Human IT (1000 visually most responsive voxels)
Human Early Visual Cortex (1057 visually most responsive voxels)
Human • fMRI in four subjects
(repeated sessions, >12 runs per subject)
• "quick" event-related design (stimulus duration: 300ms, stimulus onset asynchrony: 4s)
• fixation task (with discrimination of fixation-point color changes)
• occipitotemporal measurement slab (5-cm thick)
• small voxels (1.95!1.95!2mm3) • 3T magnet, 16-channel coil
(SENSE, acc. fac. 2)
Monkey (Kiani et al. 2007) • single-cell recordings
in two monkeys • rapid serial presentation
(stimulus duration: 105ms)
• fixation task • electrodes in anterior IT
(left in monkey 1, right in monkey 2) • 674 cells total • windowed spike count
(140-ms window starting 71ms after stimulus onset)
Monkey-Human Comparison Procedure
average of 4 subjects fixation-color task 316 voxels
average of 2 monkeys fixation task
>600 cells
Boynton (2005), News & Views on Kamitani & Tong (2005) and Haynes & Rees (2005) Kamitani & Tong (2005)
Lower spatial frequency clumping
Methodology
Interpretation Applications
Technology High field strength Coil arrays High resolution Novel functional contrast
Paradigm Designs Processing Methods
Fluctuations / Correlations Dynamics Healthy Brain Organization
Focus of this lecture!
• Field Strength • Echo Time • Spin-echo vs Gradient Echo • Velocity Nulling • RF coil arrays • High Spatial Resolution
• High Temporal Resolution • Choice of Flip Angle • Choice of Slice Thickness • Paradigm Design
• Ultimate Sensitivity? • Separating “good” and “bad” signal in Resting
State fMRI. • Understanding dynamic nonlinarities • Understanding and Using fMRI Patterns
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�