Collaborators

1
Collaborators
Part 1 Extracting temporal behavior of
Neurotransmitters from PET data
Karmen Yoder, David Kareken, Ray Muzic, Jan
Froehlich, Marc Normandin, Cristian
Constantinescu, Sean OConnor, Charlie Bouman,
Mustafa Kamasak, Brad Christian, Chunzhi Wang,
Qi-Huang Zheng, Bruce Mock, Gary Hutchins, Alex
Radnovich, Regat Seyoum, Kevin Perry, Susan
Giger, Tanya Martinez, Matthew Brown, Patrick
Aitchison, Victor Vitvitskiy, Barbara
Glick-Wilson, Mike Sullivan Whitaker Foundation,
Indiana Alcohol Research Center, General Clinical
Research Center, NIH
2
Outline of Part 1

• Goal Extract temporal information from PET.
• Motivation Understanding alcohol
  consumption/craving.
• Method - Combined modeling of tracer binding and
  competition with endogenous dopamine.
• Results - i. Simulations, ii. Rat scan data

3
What aspect of neurochemistry interests us?The
speed of the dopaminergic response to stimuli,
for one.Why?
4
Volkow review paper
DA Conc. (nM)
Time (sec)
5
Where shall we look for this response?
6
cues activate NucAcc
7
A. B.
HR AROS gt NAROS HR gt LR, AROS gt NAROS
Effects of alcohol-related olfactory stimuli in
10 subjects at behavioral (high recent drinking)
and genetic (2 family members with alcoholism)
risk for alcoholism. A) Top AROS gt NAROS effects
in nucleus accumbens (NAc). Bottom AROS gt NAROS
effect in the right NAc as compared to 5 low-risk
social drinkers (low recent drinking, no family
history). B) Top AROS gt NAROS effect in the
ventral tegmental area (VTA) in high-risk
drinkers. Bottom AROS gt NAROS effect in the VTA
compared to 5 low-risk controls.
Kareken, et al., 2004, ACER
8
But that was fMRI.What can we say about
dopamine by using PET?How would one go about
imaging dopamine?
9
Basis for Neurotransmitter Imaging?
Competition!
10
Basis of ntPET Imaging?
Competition!
Analogy BOLD effect is the basis for fMRI
11
A Model of competition must include the
Endogenous Species
voxel
Plasma
Bound
Free
Non-specifc
voxel
12
How to Include Endogenous Effects in the Model
Equations?
rate constant available receptors
radioligand
New expression for available receptors.
13
Plasma input can be expressed as a function of
PET measurement in a reference region.
FR and dFR/dt are measured
14
ntPET Estimates DA(t)
voxel
Plasma
Bound
Free
Non-specifc
Propose a functional form for Free Dopamine
variation
DA(t) Basal Gt-tDaexp(-bt-tD)
15
Rat experiments
16
Acquire Data in Two Sessions
1
2
fit data (rest drug) to model simultaneously
17
By fit the data simultaneously I mean
Minimize a single objective function to yield the
best fit(s) of rest and activated state
data, where the parameters that describe tracer
kinetics, Qexo, are the same in the R and A
states, but the parameters that describe
dopamine, Qendo, are only allowed to produce
dopamine transients above baseline in the A state.
18
Iterative Scheme for DA(t) estimation
Measured PET
Modeled Dopamine
REST
DRUG
19
Typical Rat Data from IndyPET II scanner with
11C-raclopride(2D mode direct and cross planes
only)
Not much spatial resolution But there is
molecular resolution Is there temporal
resolution with respect to neurotransmitter
change?
20
Simulated data DA timing recovery
DA peak found within 20 seconds of true time
no consistent DA peak from noise alone
small faux-DA peak w/ extreme DBF
Morris et al., ntPET A New Application of PET
Imaging for Characterizing the Kinetics of
Endogenous Neurotransmitter Release Molecular
Imaging, 2005.
21
Simulated DA activation
22
Simulated null case
Detectability threshold
23
False Positives
Based on analyses of many simulated null cases
(at the noise level of our rat data), we find
false positive dopamine responses (above
detectability threshold) 1/10 times for a
window from 5-50 min. 1/20 times for a window
from 4-45 min.
24
Simulated null with decrease in K1
25
Rat Experiments DA timing recovery
DA peak 30 min after IP alcohol 1g/kg
no DA peak with no alcohol
no DA peak 150 min after IP alcohol
Morris et al, Molecular Imaging, 2005
26
Alcohol Experiment
27
Control Experiment
28
Alcohol 2.5 hours pre-scan
29
Control Analysis
Analyze the rest case as the alcohol scan and
vice versa.
forward analysispeak time   20.9 /-
0.77delay time    9.7 /- 4.22peak height  
3.6 /- 1.45
reverse analysispeak time     8.9 /-
12.8delay time  -1.5 /- 12.5peak height  
1.0 /- 0.05
30
Are preliminary ntPET findings consistent with
microdialysis data?
Comparison with microdialysis data from 21 rats
in Heidbreder and deWitte, Pharmacol Biochem
Behav, 1993
31
Data FittingThis is a difficult problem of
limited parameter identifiability.What tricks
can we play to improve our chances of success?
32
Constrained Optimization via use of Prior
Knowledge
Objective function yields best fit of rest and
activated state data.
Constrained F yields best fit that is consistent
with prior data.
Constrained F yields best fit that is far from
practical limits.