Please enable JavaScript.
Coggle requires JavaScript to display documents.
fMRI IN TRANSLATION: THE CHALLENGES FACING REAL-WORLD APPLICATIONS -…
fMRI IN TRANSLATION: THE CHALLENGES FACING REAL-WORLD APPLICATIONS
1) INTRODUCTION
Functional neuroimaging had a big impact in brain research, but with less successes in
translation
(especially in the clinical field)
2) MEASUREMENT AND ANALYSIS
2.1) WHAT DOES fMRI MEASURE?
Features of fMRI
Good spatial resolution without constrast agents
Non invasive technique but we have to garantee safety
From what?
Heating of tissue cause electromagnetic radiation
Induction of currents
Exposure to noise
Meansures blood oxygen-level-dependent (
BOLD
signal) associated with deoxy-haemoglobin
From this, it was created the
haemodynamic response function
(HRF)
Two possible explanations for its latter part
Oxygen overcompensation as a by-product of glucose metabolism
BOLD response as indirect lined to neural activity and reliable meansure
Confirmed by Logothetis et al. study
Oxygen overcompensation as derived from neurons' need for oxygen
2.2) HOW DO WE GET TO BLOBS ON BRAINS?
Initially, fMRI datas (3D voxels) are manipulated prior analysis by
pre-processing
Datas realigned, co-registered, spatially-normalised, smoothed
The datas receive the
subject-level analysis
to create a design matrix
Onsets of each data is matched with average HRF (
convolution
) to create the
regressors
of design matrix
This explains the
low temporal resolution
Then we design sensitive
scanning experiments
(or block design or event-related designs)
Depends if timescale of alternating stimuli/tasks is similar or not to HRF
We analyze the datas with the
standard fMRI analysis
(mass-univariate approach) the magnitude of effect of each regressor
We obtain the
beta value
(slope of regression line) and this parameter is compared to create
constrast estimates
Then we conduct
group-level analysis
for every voxel of whole brain
Statistical images are overlaid with anatomical scan, so we have the presentation of
blobs on brains
3) CHALLENGES IN INTERPRETING FMRI DATA
3.1) HOW NEURAL IS THE BOLD SIGNAL?
BOLD signal
is
not so clear
(for example not only reflects neural activation)
Possible interpretation for it:
Reflects synaptic activity (common)
Removal of lactate, adjustment of acid-base and ionic balance, temperature regulation (new)
3.2) EFFECTS OF MEDICATION
Medication can change BOLD signal because they
affect
brain's
vasculature directly
(ex. serotonin)
Possible solutions:
Test unmedicated patients
Selection bias :(
Administer medication to healthy volunteer subjects
Ethical issues :(
Estimate BOLD indipendent of neural activity
Not completely solve the problem :(
3.3) MULTIPLE COMPARISONS AND EFFECT SIZES
fMRI determines an
enormous amount of data
collected and we are interested in their
effect size
In analyzing these datas we choose the
peak voxel
(value with maximal signal) and so we
overstimate
the effect of size (methodological problem)
Resolution
not trivial (and this is relevant for real-world applications)
4) GROUPS VS INDIVIDUALS
4.1) DISCRIMINATION
Respect neurology practice,
psychiatric diagnosis
with fMRI don't have high discriminability
These are
not recommended
cause expensive and potentially burdening
4.2) ANATOMICAL VARIABILITY
Necessity to understand brain structure and its function (depends on
brain's functional specialisation
of a specific area)
Trying to
reduce
the number of
cognitive processes
associated with areas' activation
Every
brain
is
unique
even our classification based on
coarse-grained labels
(thanks to which standardised our datas in pre-processing)
Trying to improve our anatomical localisation (considering inter-subject variability)
We know less about variability in functionality
Using common references like Brodmann that are not representative of humans)
5) CONCLUSIONS
Be cautious in
clinical domain
(such experiments with depressed patients are anti-depressants
Some limitations:
Limitated number of studies
Variable results
Lack of independent validation experiments
In the field of
fMRI-based lie detection
be careful for dangers in classifying erroneously
In
neuromarketing
and
neuropolitical
no problems with transferability but be caregour about ecological validity and problems with reverse inference scheme
Importance of understanding better limitations and possibilities of fMRI studies