Anda di halaman 1dari 22

Comparison of meta-analysis approaches

for neuroimaging studies of reward


processing: A case study
Manisha Chawla, Krishna Prasad Miyapuram
Cognitive Science & Computer Science
Indian Institute of Technology, Gandhinagar
Overwhelming number of studies!!

 Data Explosion
 One Participant data
 Voxel size: 3 × 3 × 5 mm
 # of slices: ~ 20
 Matrix size: 64 × 64 per slice (in-plane resolution)
 Total # of voxels (raw data) = 81920
 One study
 ~ 20 participants Number of fMRI Studies
 Multiple Studies 30000

25000
 Studies over the years !! 20000

15000

10000

5000

0
1990
1991
1992

1994
1995

1997
1998

2000
2001
2002
2003
2004
2005
2006

2008
2009

2011
2012

2014
1993

1996

1999

2007

2010

2013
Classical analysis
y = Xb + e
Statistical Analysis
fMRI General Linear Model
z time series Design matrix
Parameter estimation
roll
pitch
x Statistical Inference
Within-subject registration
slice-timing correction Linear Contrasts
Realignment Thresholding
y yaw
Coregistration
(structural to functional) Random Effects Analysis
AB (Group analysis only)
Between-subject registration
spatial normalization
Statistical
Spatial smoothing Parametric Map

Preprocessing
What happens to Variability ??
Across Participants Across Experiments
Different People Different Brains!!

Analysis Techniques

Scanner Strength
Meta Analysis
 Number of studies which look at a single psychological
phenomenon are pooled together
 analyzed for consistency and specificity of findings

Multi-level kernel Density Analysis

Image based
Intra-subject modelling
of each subject’s fMRI
time series data

Inter-subject analysis
for each study

Activation Likelihood Estimation


Kernel Density Analysis Inter-study meta-
Co-ordinate based analysis
Reverse Inferencing in single studies

P(Activation|Cognitive State)

P(Cognitive State|Activation)
Stages of Electronic Data Capture
Brainmap.org
Meta analysis of reward processing studies
Prefrontal Striatum
Cortex

Substantia
Nigra

Orbitofrontal cortex

Ventral Tegmental Area


A case study of reward-related studies
ALE analysis

MKDA

NeuroSynth forward inferencing

NeuroSynth reverse inferencing


0.00 0.20 0.40 0.60 0.80 1.00 0.00 0.20 0.40 0.60 0.80 1.00

Pallidum Pallidum
Caudate Caudate
Putamen Putamen
Amygdala Amygdala
Cingulum_Ant Cingulum_Ant
Olfactory Olfactory
Thalamus Thalamus
Frontal_Med_Orb Frontal_Med_Orb
Insula Insula
Cingulum_Mid Cingulum_Mid
Cingulum_Post Cingulum_Post
Hippocampus Hippocampus
Supp_Motor_Area Supp_Motor_Area
Rectus Rectus
Frontal_Inf_Oper Frontal_Inf_Oper
Frontal_Inf_Tri Frontal_Inf_Tri
Frontal_Sup_Medial Frontal_Sup_Medial
Parietal_Inf Parietal_Inf
Frontal_Inf_Orb Frontal_Inf_Orb
Frontal_Mid Frontal_Mid
Frontal_Mid_Orb Frontal_Mid_Orb
Parietal_Sup Parietal_Sup
Lingual Lingual
Angular Angular
Precentral Precentral
ParaHippocampal ParaHippocampal
Occipital_Sup Occipital_Sup
Occipital_Inf Occipital_Inf
Frontal_Sup_Orb Frontal_Sup_Orb
Fusiform Fusiform
Rolandic_Oper Rolandic_Oper
Precuneus
ALE_only Precuneus Forward_only
Vermis_4_5 both Vermis_4_5
both
Calcarine Calcarine
Occipital_Mid MKDA_only Occipital_Mid Reverse_only
Cerebelum_6 Cerebelum_6
Postcentral Postcentral
Customised Meta Analysis pipeline

Search Term

MATLAB Code for PubMed Interface

Search URL

Search Results
(PMID, Title, Authors, Abstract, Citations)
Decision Making in different domains
Perceptual Decision Making Value-based Decision Making

What do you prefer?

Social Decision Making


Utimatum Game
$10 Accept

$2

$0
Reject
Keywords

 508 studies identified with the terms


 reward, decision, choice, value, social, percept
 Filtering criteria:
 value: 84
 social: 61
 perceptual: 36
Perceptual Decision making

Value-based decision making

Social decision making


Data Mining: Steps

1. Database (pubmed ids) extracted from http://Neurosynth.org


(February 2014)
2. Title and abstract extracted from pubmed database (8061 studies)
3. 2518 studies that contained at least one of the keywords – decision,
value, social, percept, choice (and corresponding word forms, chose,
choose)
 (2718 studies including the additional keyword reward)
4. 639 studies had necessarily contained the substring decision or choice
(chose, choose) – final subset
Analysis: Steps

 3 separate meta-analyses were carried out for value-based,


perceptual, and social decision making, using GingerALE
 These were further entered into a pair-wise contrast analyses
 Each meta- analysis contrasts included the conjunction of
activations and
 two directional contrasts for each pair
 Contrast maps were labeled with the help of WFU-pickatlas
an Automated Anatomic Labeling software
 We calculated the number of supra-threshold voxels and the
proportion of each brain region activated (Using labeling
procedure)
Results Percept Value Social

 Common neural activation across all three pairs of


 primarily in basal ganglia (Right Caudate, Bil. Putamen, Bil. Pallidum) and Bil.
Insula, Left Inferior frontal area, Supplementary Motor Area, Right Thalamus.

 Conjunction Analysis
 Perceptual and social decision making was found to activate Anterior Cingulate
Cortex,
 Value and perceptual decision making activated Left Inferior parietal area, and
 Social and value based decision making activated left Angular gyrus, Left
Caudate, Anterior Cingulate Cortex, medial orbitofrontal and superior medial
frontal areas.
 Suggests common brain areas in pair-wise combinations of domains of
decision making.
Results Percept Value Social

 Pairwise comparison contrasts revealed the following Brain activations


 Inferior frontal areas were more active in perceptual dm vs.value-based and
social decision making
PerceptVsSocial

medial ofc, superior medial frontal, anterior cingulate, bil precentral/ inf frontal, inf parietal
 Social dm vs. Perceptual dm, greater activity in Anterior Cingulate and
medial prefrontal areas (medial orbitofrontal, superior medial frontal)
PerceptVsValue

medial ofc, caudate, ant cingulate, post cingulate, inf frontal, inf parietal
SocialVsValue
 Bil. Precentral gyri, and Inferior parietal areas to be more active in perceptual dm
vs. value-based and social dm
Results
PerceptVsValue

 Right Insula was more active in the contrast Social compared to Value-
based decision making
medial ofc, caudate, ant cingulate, post cingulate, inf frontal, inf parietal
SocialVsValue

Right Insula, caudate, putamen, pallidum, mid cingulate

Further Steps
 Reward vs decision studies
 Choice vs decision studies
 Rest studies vs common decision making studies
 Value, social, perceptual studies without decision making
References

 Dolan, R. J. (2002). Emotion, cognition, and


behaviour. Science, 298(5596), 1191-1194
 Wager, T. D. (2005). The neural bases of placebo effects in anticipation
and pain. In Seminars in pain medicine (Vol. 3, No. 1, pp. 22-30). WB
Saunders.
 Nee, D. E., Brown, J. W., Askren, M. K., Berman, M. G., Demiralp, E.,
Krawitz, A., & Jonides, J. (2013). A meta-analysis of executive
components of working memory. Cerebral Cortex, 23(2), 264-282.
 Yarkoni, T., Poldrack, R. A., Nichols, T. E., Van Essen, D. C., & Wager, T.
D. (2011). Large-scale automated synthesis of human functional
neuroimaging data. Nature Methods, 8(8), 665-670.
 Salimi-Khorshidi, G., Smith, S. M., Keltner, J. R., Wager, T. D., &
Nichols, T. E. (2009). Meta-analysis of neuroimaging data: a comparison
of image-based and coordinate-based pooling of
studies. Neuroimage, 45(3), 810-823.

Anda mungkin juga menyukai