Resting-state fMRI: Algorithms and Applications to Brain Disorders

1

Resting-state fMRI:Algorithms and

Applications to Brain Disorders

ZANG Yu-Feng 臧玉峰 M.D. [email protected]

YAN Chao-Gan 严超赣 Ph. [email protected]

State Key Laboratory of Cognitive Neuroscience and Learning,

Beijing Normal University, China

2

Outline

• What is resting-state fMRI?

• Computational methodology

• Applications to brain disorders

3

Functional MRI?

Generally

• Blood oxygenation level dependent (BOLD)

• Perfusion

• Diffusion

• Magnetic resonance spectroscopy (MRS)

• ……

4

Functional MRI?

More Specifically

• BOLD-fMRI

• Perfusion

Most Specifically• BOLD-fMRI

5

Functional MRI?

6

Task-state fMRI?

T C T C T C Design

Expected signal

Contrast within a scanning session is necessary!

7

Task-state fMRI

baseline or control

activated

delta

What is activation?

8

Task-state fMRI

Abnormal baseline?

Abnormal activated level?

Abnormal delta

What is abnormal activation for patients?

So, baseline is important!

9

What is resting-state fMRI?

• Eyes closed or open with no task

• Continuously

• Quite a few minutes or more

• Not to fall asleep

• Not to think of anything in particular

• Low frequency fluctuation (LFF, 0.01 – 0.08 Hz)

10

The nature of task BOLD

“the haemodynamic response seems to be better corr

elated with the LFPs, implying that activation in an a

rea is often likely to reflect the incoming input and th

e local processing in a given area rather than the spik

ing activity.”

(Logothetis et al., 2001, Nature)

BOLD ≈ LFP

11

The nature of spontaneous BOLD

“It was found that the impulse response function comp

uted from LFP and BOLD under conditions of no sti

mulation can predict the response under stimulation

conditions.”

(Logothetis et al., 2001, Nature)

Spontaneous BOLD ≈ spontaneous LFP

12

Resting-state BOLD-fMRI signal reflects:

However, also physiological noise (respiration, cardiovascular pulsation, etc) (Cordes et al., 2001; Birn et al., 2006)

13

Outline



• Applications to AD studies

14

Task-state fMRI

• Functional integration

(inter-regional relationship or connectivity)

• Functional segregation

(activation detection)

1 >> 2

15

Most of resting-state fMRI studies： integration (connectivity)

Correlation: (Biswal et al., 1995; ……)

ICA: (Kiviniemi et al., 2003; van de Ven et al., 2004; G

reicius et al., 2004)

Hierarchical Clustering: (Cordes et al., 2000; Salva

dor et al., 2005)

Self Organization Map: (Peltier et al., 2003)

....

16

The first Resting-state fMRI study

(Biswal et al., 1995)

Bilateral finger tapping (task vs. rest)

Resting-state correlation of low frequency fluctuation (LFF, 0.01 – 0.08 Hz)

(Courtesy of Dr. WENG Xu-Chu)

17

Linear Correlation

Spontaneous LFF was highly synchronous among:

Bilateral motor cortices (Biswal et al., 1995)

Bilateral visual cortices (Lowe et al., 1998;

Kiviniemi et al., 2004)

Bilateral auditory cortices (Cordes et al., 2001)

Bilateral amygdala (Lowe et al., 1998)

Bilateral thalamus (Stein et al., 2000)

Language cortices (Hampson et al., 2002)

Default mode network (Greicius et al., 2003; Fox et al., 2005..)

18

A default mode of brain function(Raichle et al., 2001)

PCC, MPFC, IPL etc., 1) Consistently show task-independent deactivation

(TID) during a wide range of tasks with cognitive demand

2) Highest metabolic rate in the brain during resting state

1)

2)

19

Functions of the default mode network

Intrinsic activities (Fransson, 2005):

• Self-reflection, self-referential processing, self awareness

• Task-unrelated thought (TUT)

• Mind wandering or daydreaming

• Streams of consciousness

• Memory retrieval of autobiographic events

• ……

20

Most of resting-state fMRI studies： integration (connectivity)

Correlation: (Biswal et al., 1995; ……)

ICA: (Kiviniemi et al., 2003; van de Ven et al., 2004; G

reicius et al., 2004)

Hierarchical Cluster: (Cordes et al., 2000; Salvador

et al., 2005)

Self Organization Map: (Peltier et al., 2003)

....

21

Independent Component Analysis (ICA)

Spatial ICA for fMRI data: separate a mixture of a 4D data into a set of spatially independent components (McKowen et al., 1998)

22

Spatial ICA for resting-state fMRI

Vascular component Visual component

(Kiviniemi et al., 2000)

23

“Integrative” is really good, but:

Decreased

functional connectivity

Question: Is A, B, C, or……abnormal?

24

Only A few resting-state fMRI studies：Segregation

rms, power spectrum, ALFF: (Biswal et al., 1995; Li et al., 2000; Kiviniemi et al., 2000; Zang et al., 2007)

TCA: (Liu et al., 2000; Morgan et al., 2004)

Regional Homogeneity: (Zang et al., 2004)

Multiple Regressors: (Fransson, 2005)

Autoregressive Noise Model: (Cordes et al., 2005) Fractional Gaussian Noise: (Maxim et al., 2005)

25

rms, power, ALFF

For a given frequency:

root mean square (rms)

standard deviation

amplitude

rms: white matter vs. gray matter = 0.6 : 1 (Biswal et al., 1995; Li et al., 2000)

Square root of the power

26

rms, power, ALFF

Power spectrum:

Higher power at 0.033Hz in visual area (Kiviniemi et al., 2000)

27

Amplitude of Low Frequency Fluctuation

(ALFF) (Zang et al., 2007; Yang et al., 2007)

0.01-0.08 Hz

TR

400 ms

1.25 Hz

Steps: square root, average of 0.01-0.08 Hz, standardization by global mean

28

ALFF

ALFF(Zang et al., 2007)

PET(Raichle et al., 2001)

noise

29

Abnormal ALFF

ADHD vs. controls

(n=13) (n=12)

(Zang et al., 2007)

Early AD vs. controls

(He Y et al., 2007) (Please see below for details)

30

Improvement: fractional ALFF

Suprasellar cistern

PCC

31

Improvement: fractional ALFF

(Zou et al.,

2008)

32

Regional Homogeneity (ReHo)

Similarity or coherence of the time courses within a functional cluster

(Zang et al., 2004)

33

ReHo: motor task state vs. pure resting state (Zang et al., 2004)

Rest > Motor

Motor > Rest

a) Higher ReHo in bilateral primary motor cortices during motor task

b) Higher ReHo in default mode network (PCC, MPFC, IPL) during rest (Raichle et al., 2001; Greicius et al., 2003)

34

ReHo applications to brain disorders

ADHD: Zhu et al., 2005, 2007; Cao et al., 2006

AD/MCI: He et al., 2007; Bai et al., 2008

Schizophrenia: Liu et al., 2006, Shi et al., 2007

Aging: Wu et al., 2007

PD: Wu et al., 2008

Depression: Yuan et al., 2008

35

Outline




36

RESTing-state fMRI data analysis toolkit (REST)

By Song et al., http://www.restfmri.net

37

Data Processing

Assistant for Resting-State

fMRI (DPARSF)

Yan CG* and Zang YF* (2010) DPARSF: a MATLAB toolbox for "pipeline" data analysis of resting-state fMRI. Frontiers in Systems Neuroscience. 4(13) http://www.restfmri.net

http://www.restfmri.net/



38

Outline




39

Some issues about resting-state fMRI

• Definition and instruction?

• Sampling rates (TR)

• Length

• Standard method for data analysis?

40

Definition and Instruction?

• Eyes closed or open with no cognitive task

• Continuously

• Not to fall asleep

• Not to think of anything particular

ill defined!

41

Different resting conditions?

The functional connectivity patterns of the default mode network was minimally disturbed during different resting conditions with limited cognitive demand (Greicius et al., 2003; Fransson, 2005; Fox et al., 2005)

Fox et al., 2005

42


Significantly different between different resting conditions with limited cognitive demand Yan et al., PLoS ONE 2009

Functional connectivity maps

43


Significantly different between different resting conditions with limited cognitive demand

Regional activity (ALFF)

Yan et al., PLoS ONE 2009

44


Suggestion: eyes closed for patients studies

45

Is the resting block a resting-state?

In block design

Long resting

No study has compared directly between them yet

46

Is the resting block a resting-state?

Stimulus

Peak: 4-8 s

BOLD

Post undershoot: > 15 s

20 - 30 s

47




• Length

• Standard method for data analysis?

48

Sampling rates (TR)

Nyquist sampling theorem:

Shorter TR (<400 ms, >2.5 Hz)

for heart beating (around 1.3 Hz)


49

But for respiration (around 0.3 Hz)?

Is TR <1.5 s fast enough?

Simultaneous recording of respiration and heart beat to remove physiological noise (Hu et al., 1995, MRM)

Courtesy of SONG Xiao-Wei

Sampling rate and physiological noise

50



• Length

• Standard data analysis?


51

Length?

LFF: 0.01-0.08 Hz

Peak around 0.033 Hz

(Kiviniemi et al., 2000, MRM)

≈2/min


8 min, we used

52



• Length

• Standard data analysis?

Not yet


53

Applications to brain disorders

• Functional connectivity (integration)

Linear correlation

Independent component analysis

• Regional activity (functional segregation)


Amplitude of low frequency fluctuation (ALFF)

54

Functional connectivity, linear correlation analysis

Multiple Sclerosis: Low-Frequency Temporal

Blood Oxygen Level–Dependent Fluctuations

Indicate Reduced Functional Connectivity -

Initial Results

(Lowe et al., 2002)

55

Data analysis and results

MT: the middle part of corpus callosum (connect

ing bilateral M1), lower in MS patients

(Lowe et al., 2002)


56

Decreased functional connectivity between bilatera

l sensorimotor cortex

(Lowe et al., 2002)


57

Changes in hippocampal connectivity in the early stages of Alzheimer's disease: ev

idence from resting state fMRI

(Wang L et al., 2006)

• Pure resting state

• Subject-specific Seed ROIs in bilateral hippocampus

• Correlation analysis


58

(Wang L et al., 2006)

AD: decreased connectivity with the right hippocampus

59



Linear correlation





60

Default-mode network activity distinguishes Alzheimer's disease from healthy aging: evidence from functional

MRI

(Greicius et al., 2004)

Functional connectivity, Independent component analysis

(ICA)

61

Methods

Data: sensorimotor task in event-related design

Independent component analysis (ICA)Define the default mode network componen

t by “goodness-of-fit” with a PCC-templateBetween-group comparison on the Z mapsBetween-group comparison on the “goodnes

s-of-fit”


62

Increased functional connectivity in normal elderly

“Goodness-of-fit”


63



Linear correlation





64

Regional coherence changes in the early stages of Alzheimer's disease: a

combined structural and resting-state functional MRI study

(He Y et al., 2007)

• Pure resting state• Regional homogeneity (ReHo)• Amplitude of low frequency fluctuation (ALFF)

65

MCI: decreased ReHo and ALFF in PCC, while increased in visual cortex (p < 0.001, corrected)

(He Y et al., 2007)

66

Consistent with PET studies (Reiman et al., 1996; Minoshima et al., 1994)

Correlated with MMSE

(He Y et al., 2007)

67

Summary on the resting-state fMRI

Mechanism: needs further studies Application to clinical studies:

easy, cheap Combine with other MRI: structure, per

fusion (ASL), DTI

68

Multiple MRI modalities

DTIfMRI: Task, Rest

MRS

Others

Computational Neuroanatomy

69

Multiple Disciplines

Neurology and PsychiatryNeuroradiologyComputer ScienceCognitive Neuroscience……

70

Thanks to

拼音顺序 :贺永贾建平金真蒋田仔李坤成Kiviniemi, Vesa杨一鸿王玉凤翁旭初谢晟张岱朱朝喆……

All the group members!

NSFC

30470575 30625024

30728017 30621130074

71

Thanks for your attention!

Documents

Resting-state fMRI: Algorithms and Applications to Brain Disorders