fMRI: Interpretation, Limits and Potential Pitfalls - … · fMRI: Interpretation, Limits and Potential Pitfalls Seong-Gi Kim [email protected] . Mapping Brain Functions ... Duvernoy

fMRI: Interpretation, Limits and Potential Pitfalls

Seong-Gi Kim

[email protected]

Mapping Brain Functions

Neural Activity Vascular Response

Stimulation/Task Functional Map

(MRI)

Pre-synaptic activityPost-synaptic activityAction potentials

Blood flowBlood volumeBlood oxygenation

Blood oxygenation level~1.0

~0.6

CapillariesArteries Veins

Vascular Structure

Distance

Blood Oxygenation Level-dependent ContrastOgawa et al. Magn. Reson. Med, 1990

HEMOGLOBIN

Deoxyhemoglobin (Paramagnetic)

Oxyhemoglobin (Diamagnetic)

-O2

Reduce T2* -> Reduce signal intensity in T2*-weighted images

4 mm

T2*-weighted images of rat brain (no activation)(isotropic resolution of 58 μm, 9.4 T)

Dark lines – venous vessels (>20 micrometer diameters)

Sung-Hong Park et al., Magn. Reson. Med., 2008

Blood Oxygenation Level-dependent Contrast

Mouse brain images at 360 MHz

Breathing 100% O2Breathing air

Ogawa et al. MRM, 1990

Dynamic BOLD MR Measurements in Cats

Turner R, Le Bihan D, Moonen CT, Despres D, Frank J “Echo-planar time course MRI of cat brain oxygenation changes” Magn Reson Med. 1991 Nov;22(1):159-66

Abstract: When deoxygenated, blood behaves as an effective susceptibility contrast agent. Changes in brain oxygenation can be monitored using gradient-echo echo-planar imaging. With this technique, difference images also demonstrate that blood oxygenation is increased during periods of recovery from respiratory challenge.

Blood oxygenation level~1.0

~0.6

CapillariesArteries Veins

Vascular Structure

Distance

(Task -> oxygen supply overcompensatesoxygen utilization)

(Fox et al., 1988)

University of Minnesota/Bell LabOgawa et al. Proc Natl Acad Sci USA, 1992

Functional Image(Visual Stimulation)

Anatomical Image

One of First Human fMRI StudiesPrimary Visual Cortex

- Functional MRI with a few millimeter resolution is routinely used for mapping brain functions such as vision, motor, language, cognition, etc.

- Underlying assumption is that fMRI signal change is indirectly related to neural activity, and its location is indicative of neural activity site.

Current Status of Functional MRI

Physiological Changes

Biophysical Basis of BOLD fMRI

Spatial Resolution

Interpretation - Quantification

Temporal Resolution

Vascular Physiological Changes

Costantino Iadecola & Maiken NedergaardNature Neurosci, 2007

Blood Vessel Dilation

Blood Velocity Increase

Blood Oxygenation Change

Cerebral Blood Flow

Vessel Imaging of Rat Brain

Anatomical Image

1 mm

Hig

h-re

solu

tion

Ana

tom

ical

Imag

e 200µm

Vazquez et al., 500

µm

200µm

20x Mag., Reverse contrast

1 mm

Vazquez et al.

Bright:

dilation

A

V V

Vessel Imaging of Rat Brain during Stimulation

4-s forepaw stim

QUANTIFICATION (PO2)

ArteriesVeins

SAMA

LA

SVMV

LV

T

LDF(CBF)LDF(CBF)

TissuePO2

TissuePO2

VenousPO2

VenousPO2

Vazquez et al., JCBFM, 2010

Simultaneous measurements of CBF and PO2

Clark-type oxygen sensor (30 and 4 μm diameter)

CBF and tissue PO2 changes during stimulation

Forepaw Stimulation

Time (s)

LDF (CBF)

Tissue PO2

Vazquez et al., JCBFM, 2010

Lar. Ven.Lar. Ven.

Med. Ven.Med. Ven.

Sm. Ven.Sm. Ven.

Venous Blood PO2 changes during stimulation

Vazquez et al., JCBFM, 2010S

O2

PO2

Physiological Changes

Biophysical Basis of BOLD fMRI

Spatial Resolution

Interpretation - Quantification

Temporal Resolution

Intravascular water moves freely

Slow exchange of IV and EV waterIntact BBB ⇒ tight junctions betweenendothelial cells impede the diffusion of water. (In 50 ms, less than 5% of the capillary water diffuses into the EVS.)

Extravascular water moves freely

EV

RBC

blood vessel(IV)

Compartmentalization of Water

Reb Blood Cell

water

Intravascular Effect -> T2 Change

Water appears to move freely across the RBC membrane(residence time in RBC ~ 5 ms).

Susceptibility effect in Extravascular Pool:

1% of maxat r = 10 a.

30 μm 300 μm

Δωout = Δωmax (radius of vessel/distance from vessel)2

Related to deoxyhemoglobin concentration (oxygen saturation level) × magnetic field strength

Susceptibility effect in Extravascular Pool:

1% of maxat r = 10 a.

30 μm 300 μm

ΔBoutMRI signal at echo time (TE):

a summation of all water proton signal withina voxel. Each proton signal decays by T2 and dephases by local susceptibility effect (i.e., Phase shift)

S(TE) = ∑ S.exp(-TE/T2).e(-iϖTE)

≈ S.exp(-TE/T2*)

x’

y’

x’

y’

x’

y’

x’

y’

t = 0 (after 90° pulse)

τ

τ

180° pulse along x’

Spin-echo

Spin Echo (two spins)

Capillary tube (1.4 mm o.d., 1.0 mm i.d.) filled with blood in a saline bath(positioned orthogonal to main magnetic field)

100% oxyHb

SE GE

100% deoxyHb

Ogawa et al., MRM, 1990

Boxerman et al., MRM, 1995

CBV = 2%Δχ = 0.1 ppm

Conventional Gradient-echo and Spin-echo BOLD Signal

Extravascular and Intravascular BOLD Signal Contributions

Gradient Echo Spin Echo

X

X

X

X

X

X

X

Extravascular

Largevessels

Smallvessels

IntravascularLarge

Small

Physiological Changes

Biophysical Basis of BOLD fMRI

Spatial Resolution

Interpretation - Quantification

Temporal Resolution

Vascular Structures - Histology

Midline Human visual cortex

White matter~ 2 mm

Duvernoy et al.Brain Research Bulletin, 1981

Since all fMRI techniques rely on blood signals, it is desirable to detect responses of small vessels which are close to active neurons.

Layer 4 is known to be highest capillary density and metabolic responses.

Cortical cytoarchitectureof cat visual area 18

Cortical Layer Model

GrayMatter

PiaMatter

WhiteMatter

Vascular structure

Pia1

2

3

4

5

6

white

2 mmL

D1

23

45

6

WM

-Timan et al., Brain Res, 2004

- Torre et al., Anat Rec 1998

Cortical Depth-Dependent Gradient-echo BOLD fMRI

(156 x 156 µm2 in-resolution, 4-shot EPI, 9.4T)

0

1

2

3

4

5

6

-0.5 0 0.5 1 1.5 2 2.5

Distance from Cortical Surface (mm)

% C

hang

e (T

E=2

0 m

s, 9.

4T) WMGM

2 mm Zhao et al., NeuroImage, 2006

Scanning Electron Microscopy

500 µm

Gray Gray mattermatter

PiaPiamatermater Torre et al., 1998

(Human cortex)

Vascular structures vs. fMRI resolution

BOLD Signal ∝

Δdeoxyhemoglobin conc. in blood xvenous blood volume

0.3 3.0 (%)0.3 3.0 (%)

2 mm

TE=20 ms TE=40 ms

Gradient-echo vs. Spin-echo BOLD fMRI

(156 x 156 µm2 in-resolution, 4-shot EPI, 9.4T)

Gradient-Echo Spin-Echo

Spatial Specificity of BOLD Signal to Neural Activity Site

- Venous Vascular Structures

Pial Venous Vessels: 130 – 380 μm diameter

Intracortical Veins: 80 – 120 μm average diameter1 – 2 mm apart

Distance between Intracortical Veins

Distance between emerging venous veins: 0.75 - 4 mm

Duvernoy et al. Brain Research Bulletin, 1981

Pia

GM

WM

artery vein

Hubel & Wiesel, 1968

Ocular DominanceColumns

Color sensitiveregions

Orientation columns

Gray matter (1.5 – 3 mm)

Neurons with similar properties are clustered as columnsSingle-neuron Activities

Can you map cortical columns?

+10-10 0Signal intensity (arbitrary unit)

A P

D

V SPL

5 mm

Iso-orientation maps in the medial area using fMRI

Fukuda et al., J of Neurosci, 2006

(with contrast agent, dilation of small vessels)

A P

D

V SPL

180°

0°

90°

5 mm

Observation of orientation preference maps

Fukuda et al., J of Neurosci, 2006

Left Right

Lateral sulcus (LS)

Marginal gyrus (mg)

5 mm

LS

mgCoronal planeCoronal plane

WM

0o 90o

0.3

0.8

1 mm

BOLD fMRI

(Kim et al. Nature Neurosci, 3: 164-169, 2000)

5 mm

1.0

-1.0

ΔS

(x m

ean)

CBVGE BOLD SE BOLD

A

R

BOLD vs. CBV is-orientation maps(obtained with the differential approach; 0° – 90°)

GE CBV

2 mm

SE

Moon et al., J of Neurosci, 2007

Physiological Basis

Biophysical Basis

Spatial Resolution

Interpretation - Quantification

Temporal Resolution

BOLD Signals

Dependent on Bo, TE, pulse sequence (GE vs SE)

Dependent on vessel size, orientation, and density

Dependent on hematocrit level

Dependent on oxygenation level

ΔR2* = Δ(1/T2*) = −percent change/TE ∝ −CBVv(1 – ΔSv) + ΔCBVv(1 – Sv)

1 - Sv = CMRO2 / CBF

• Cerebral Oxygen Consumption Rate• Cerebral Blood Flow• Venous Blood Volume

where

Parenchymal Microvessel (<50 μm diameter)

Region Blood volume

Occipital cortex 1%

Corpus callosum 0.4%

Cerebellar nuclei 1.3%

Rat; Fenstermacher et al.

Task/stimulation

Neural activity

CBF CBVvCMRo2

dHb

T2* T2*

BOLD Signal BOLD Signal

rCB

V(a

rbitr

ary

unit)

rCBV = 0.31 rCBF + 0.67( r = 0.85 )

rCBF (arbitrary unit)1.0 1.4 1.8 2.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

CBV vs. CBF during Hypercapnia

58 ml/100 g/min

(α-chloralose anesthetized rats)

rCBV = 0.975rCBF0.40

(100% CBF -> 31% CBV)

Lee et al., MRM, 2001

0.6 0.8 1.0 1.2 1.4 1.6 1.80.6

0.8

1.0

1.2

1.4

1.6

1.8

rCB

V(a

rbitr

ary

unit)

rCBV (vein)

rCBV (artery)

rCBF (arbitrary unit)

CBF vs. Arterial and Venous CBV

Lee et al., MRM, 2001

Task/stimulation

Neural activity

CBF CBVCMRo2

dHb

T2* T2*

fMRI Signal Change is related to Neural Activity

LOGOTHETIS et al. Nature, 412, 150 – 157, 2001

fMRI Signal Change is related to Neural Activity

LOGOTHETIS et al. Nature, 412, 150 – 157, 2001

0.94

0.97

1

1.03

1.06

1.09

0 100 200 300 400

Time (seconds)

Nor

mal

ized

BO

LD S

igna

l

25

30

35

40

45

50

55

ETC

O2

(mm

Hg)

n = 6 subjects for each study

Visual Stimulation under different baseline conditions

hypercapnia

hypocapnia

Cohen et al. JCBFM, 2002

-2

0

2

4

6

-4 0 4 8 12 16 20 24 28 32 36 40

Time (seconds)

Ave

rage

BO

LD C

hang

e (%

)

hypocapnia

normocapniahypercapnia

-0.4

0

0.4

0.8

1.2

0 0.5 1 1.5 2 2.5 3

Time (seconds)

Aver

age

BOLD

Cha

nge

(%)

Interpretation of fMRI signals

- fMRI signal is an index of ensemble of neural activity(presumably monotonic relation)

- Difficulty to compare fMRI signals across cortical regions and subjectsdue to BOLD signal dependencies on vascular structure and volume.

- Neural source of BOLD signal is not clear – spiking activities vs. synaptic activity, excitatory vs. inhibitory

- Excellent non-invasive tool to map whole brain functions with relatively high spatial (a few millimeters in humans) and temporal resolution (~a few seconds).

Physiological Basis

Biophysical Basis

Spatial Resolution

Interpretation - Quantification

Temporal Resolution

99

100

101

102

103

104

105

0 2 4 6 8 10 12 14 16 18 20

Time (sec)

+ 2 sec

- 2 sec

0 sec

Relative Delay Time

0

10

20

30

40

50

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

Relative Delay (sec)

# of

pix

els

Provided by P.A. Bandettini

Heterogeneity of fMRI changes in humans: response times(Bilateral finger movements)

Task Execution

BOLD response

Task Execution

BOLD response

Time to peak

Inter-epoch delay time (1 – 10 sec)

2 sec

-5

0

5

10

15

20

25

0 20 40 60 80 100

time (s)

white matter

motor area

fMRI Signal vs. Finger Movements

BO

LD c

hang

e (%

)fin

ger p

ress

ure

Delay time=3 – 7.5 s

Presentation Contemplation Decision

time

displayed until decision is made

mental rotation experiment

Richter et al. J. Cogn. Neurosci, 1999

Superior Parietal Area

Supplementary Motor Area

Lateral Premotor Area

Central Sulcus

Functional Maps of Mental Rotation

-10 -5 0 5 10 15

1%R

elat

ive

fMR

I int

ensi

ty

Supplementary motor

Primary motor

Time from button press (sec)

Response Time-locked Time Courses in M1 and SMA