models of neuronal populations
DESCRIPTION
AACIMP 2010 Summer School lecture by Anton Chizhov. "Physics, Chemistry and Living Systems" stream. "Neuron-Computer Interface in Dynamic-Clamp Experiments. Models of Neuronal Populations and Visual Cortex" course. Part 2.More info at http://summerschool.ssa.org.uaTRANSCRIPT
Models of neuronal populations
Anton V. Chizhov
Ioffe Physico-Technical Institute of RAS,St.-Petersburg
Definitions:
Population is a great number of similar neurons receiving similar input
Population activity (=population firing rate) is the number of spikes per unit time per total number of neurons
Neurons
Neuronal populations
Large-scale simulations(NMM & FR-models
for EEG & MRI)
Overview
• Experimental evidences of population firing rate coding
• Conductance-based neuron model
• Probability Density Approach (PDA)
• Conductance-Based Refractory Density approach (CBRD)
- threshold neuron- t*-parameterization - Hazard-function for white noise- Hazard-function for colored noise
• Simulations of coupled populations
• Firing-Rate model
• Hierarchy of visual cortex models
• What can be modeled on population level?
• Which details are important?
• What kinds of population models do exist?
• Which one to choose?
[E.Aksay, R.Baker, H.S.Seung, D.W.Tank \\
J.Neurophysiol. 84:1035-1049, 2000] Activity of a position neuron during spontaneoussaccades and fixations in the dark. A: horizontal eye position (top 2 traces), extracellular recording (middle), and firingrate (bottom) of an area I position neuronduring a scanning pattern of horizontal eyemovements.
[R.M.Bruno, B.Sakmann // Science 312:1622-1627, 2006 ]
Population PSTH of thalamic neurons’responses to a 2-Hz sinusoidal deflection of theirrespective principal whiskers (n = 40 cells).
Commonly information is coded by firing rate
Whole-cell (WC) recording of a layer2/3 neuron of the C2 cortical barrelcolumn was performed simultaneouslywith measurement of VSD fluorescence under conventional opticsin a urethane anesthetized mouse.
Commonly populations are localized in cortical space
Voltage-sensitive Dye Optical Imaging[W.Tsau, L.Guan, J.-Y.Wu, 1999]
Pure population events observed in experiments:
• Evoked responses
• Oscillations
•Traveling waves
• What can be modeled on population level?
• Which details are important?
• What kinds of population models do exist?
• Which one to choose?
GABA-IPSC AMPA-EPSCAMPA-EPSC
AMPA-EPSP
AMPA-EPSP
GABA-IPSP
GABA-IPSC
GABA-IPSPPSP
PSP
Firing rateFiring rate
SpikeSpike
Threshold criterium
Population model
Synaptic conductance kinetics
Membraneequations
Eq. for spatial connections
Approximations forare
from [L.Graham, 1999]; IAHP is from [N.Kopell et al., 2000]
Model of a pyramidal neuron
)()()()( 0 ttuVVtsIIIIIIIdt
dVC AHPLHMADRNa η++−−−−−−−−−=
HMADRNa IIIII ,,,,
)()(
,)(
)(
U
yUy
dt
dy
UxUx
dtdx
y
x
τ
τ−=
−=
∞
∞
))(()()( ......... VtVtytxgI qp −=
Color noise model (Ornstein-Uhlenbeckprocess):
)(2 tdtd σξτηητ +−=
MODEL
EXPЕRIМЕNТ
• ionic channel kinetics• input signal is 2-d
)()()()( 0 tIVVtgtu electrodeS SS +−=∑∑=
S S tgts )()(
• synaptic channel kinetics
C
Vd
Vd
Vd
Vd
Vs
Vs
Is
Is
g=Id/(Vd-Vrev)
B
t, ms
PS
C,p
A
PS
P,m
V
0 5 10 15-200
-150
-100
-50
0
0
5
10
15
PSC, exp.PSP, exp.PSP, model 2PSP, model 1
2-comp. neuron with synaptic currents at somas
+∂
∂−−+−−−=
−−+−−=
dd
ms
drest
dd
m
s
Sd
restm
It
I
GVVVV
dt
dV
G
IVVVV
dt
dV
31
))(2()(
)()(
τρ
ρτ
ρτ
Figure Transient activation of somatic and delayedactivation of dendritic inhibitory conductances in experiment (solid lines) and in the model (small circles). A, Experimental configuration.B, Responses to alveus stimulation without (left) and with(right) somatic V-clamp. C, In a different cell, responses to dynamic current injectionin the dendrite; conductance time course (g) in green, 5-nS peak amplitude, Vrev=-85 mV.
A
[F.Pouille, M.Scanziani//Nature, 2004]
X=0 X=L
Vd
V0
Parameters of the model:τm= 33 ms, ρ = 3.5, Gs= 6 nS in B and 2.4 nS in C
Two boundary problems:A) current-clamp to register PSP: B) voltage-clamp to register PSC:
;0
∂∂+=
∂∂
= T
VVGR
X
Vs
X
)(TIRX
VS
LX
=∂∂
=
02
2
=+∂∂−
∂∂
VXV
TV
;0)0,( =TV
Solution:
• neuron is spatially distributed
[A.V.Chizhov // Biophysics 2004]
Модель. Ответ зрительной коры на полосу горизонтальной, а затем вертикальной ориентации.
Эксперимент. Зрительная кора. Картаориентационной избирательности
активности нейронов.
Модель “Pinwheels” картыориентационной
избирательности входных
сигналов.
• spatial structure of connections
1 mm
• What can be modeled on population level?
• Which details are important?
• What kinds of population models do exist?
• Which one to choose?
Population models
• Definition
A population is a set of similar neurons
receiving a common input and dispersed due to noise and intrinsic parameter distribution.
• Common assumptions:
– Input – synaptic current (+conductance)
– Infinite number of neurons
– Output – population firing rate
N
tttn
tt act
Nt
);(1limlim)(
0
∆+∆
=∞→→∆
ν
(4000)
Direct Monte-Carlo simulationof individual neurons:
Firing-rate:
Probability Density Approach (PDA):
N
ttn
tt
спайкиVVVV
tVVgIt
VC
act
resetT
ILL
)(1)(
т , если
)()(
∆+∆
=
=>
+−−=∂∂
ν
ξσ
∫∞
+
==
+=
−−=
∂∂+
∂∂
−=∂∂+
∂∂
0
*)0,()(
))/,()((1
*)),((
)(*
*
:
dtHttv
dtdUUBUAttUH
VUgItU
tU
C
Htt
модельRD
m
LL
ρρ
τ
ρρρ
))((~)(
),(
))((
))(()(
tUft
VUgIdt
dUCor
tIfdtd
or
tIft
LL
=
−−=
+−=
=
ν
νντ
ν
Types of population models
(4000)
Assumption. Neurons are de-synchronized.
f
I
“f-I-curve”
where the matrix represents the influence of noise
Problem! The equation is multi-dimensional.
Particular cases are
- membrane potential
- time passed since the last spike
- time till the next spike
Idea of Probability Density Approach (PDA)
For classical H-H:
[A.Turbin 2003]τ≡X
SXFdt
Xd rrr
r
+= )(
),( tXr
ρ
Single neuron equation (e.g. H-H model)
( )
∂∂⋅
∂∂+⋅
∂∂−=
∂∂
XW
XXF
Xtr
t
r
rr
r
ρρρ)(
Sr
Wt
),,,( nhmVX =r
Fr
where is the common deterministic part,is the noisy term.S
r
Eq. for neural density
*tX ≡
VX ≡
[B.Knight 1972]
[A.Omurtag et al. 2000][D.Nykamp, D.Tranchina 2000][N.Brunel, V.Hakim 1999], …
[J.Eggert, JL.Hemmen 2001][А.Чижов, А.Турбин 2003]
• Kolmogorov-Fokker-Planck eq. for ρ(t,V)Leaky Integrate-and-Fire (LIF) neuron:
• Refractory density ρ(t,t*) for SRM - neurons
Spike Response Model (SRM):
)'()'()(,0)(
),()(
2 ttttt
VVthenVVif
ttRIVdt
dV
m
resetT
m
−>=<>=<
=>
++−=
δστηηη
ητ
[ ] )(2
)(2
22
resetm VVV
RIVVt
−⋅+∂∂+−
∂∂=
∂∂ δνρσρρτ
Htt
ρρρ −=∂∂+
∂∂
∗ ∫∞
=≡0
** ),()()0,( dttttt ρνρ
( ) ( ) ( ) ( )∫+=*
0
*** ''',,t
dttIttktttU η
)*),,(( TVttUHH =
[W.Gerstner, W.Kistler, 2002]
TVVVt
=∂∂= ρσν
2)(
2
VTVreset
ρ
Hz
0
ν
Hz
0 t
Problem! Voltage can not uniquely characterize neuron’s state.
Simplest 1-d PDAs
1-D Refractory Density Approach for conductance-
based neurons (CBRD)
1. Threshold single-neuron model
2. Refractory density approach (t* -parameterization)
3. Hazard-function
Htt
ρρρ −=∂∂+
∂∂
∗
iAHPLHMADR IIIIIIIt
U
t
UC −−−−−−−=
∂∂+
∂∂
*
)(
)(
,)(
)(
*
*
U
yUy
t
y
t
y
U
xUx
t
x
t
x
y
x
τ
τ−=
∂∂+
∂∂
−=∂∂+
∂∂
∞
∞
t* is the time since the last spike;
[A.V.Chizhov, L.J.Graham // Phys. Rev. E 2007, 2008]
H(U) = ‘frozen stationary’ + ‘self-similar’solutions of Kolmogorov-Fokker-Planck eq. for I&F neuron with white or color noise-current
BAH +≈
∫∞
+
=0
*)( dtHtv ρ
Approximations for are taken from [L.Graham, 1999]; IAHP is from [N.Kopell et al., 2000]
1. Threshold neuron model
iAHPLHMADRNa IIIIIIIIdt
dVC −−−−−−−−=
HMADRNa IIIII ,,,,
iAHPLHMADR IIIIIIIdt
dUC −−−−−−−=
Full single neuron model
Threshold model
).1(018.0,:
);1(18.0,:
;002.0:
;691.0,743.0:
;473.0,262.0:
40
wwwwwIfor
xxxxxIfor
yyIfor
yyxxIfor
yyxxIfor
mVUUthenVUif
resetresetAHP
resetresetM
resetH
resetresetA
resetresetDR
resetT
−=∆∆+=
−=∆∆+=
==
====
====−==>
)(
)(,
)(
)(
U
yUy
dt
dy
U
xUx
dt
dx
yx ττ−=−= ∞∞
NaI−
2. Refractory density approach (t* - parameterization)
Htt
ρρρ −=∂∂+
∂∂
∗
iAHPLHMADR IIIIIIIt
U
t
UC −−−−−−−=
∂∂+
∂∂
*
)(
)(
,)(
)(
*
*
U
yUy
t
y
t
y
U
xUx
t
x
t
x
y
x
τ
τ−=
∂∂+
∂∂
−=∂∂+
∂∂
∞
∞
Boundary conditions:
-- firing rate)()0,(0
tdtFt νρρ ≡= ∫∞
+
∗
( )
)(1)exp(2
)(~,),(~2-)(
)),(),(),(),(),(/(
),()(1)(
2
m
*****
Terf
TTF
UUTTF
dt
dTUB
ttggttgttgttgttgC
dtdUUBUAUH
T
AHPLHMADRm
m
+−=−==
+++++=
+=
πστ
ττ
).),((),(:
;),()0,(
;),()0,(
;,,)0,(,)0,(
)0,(
***
*
*
dtttdUUttUt
Iforwttwtw
Iforxttxtx
IIIforytyxtx
UtU
TTTT
AHPresetT
MresetT
HADRresetreset
reset
=
∆+=
∆+=
===
),(),,(),,(),,( **** ttyyttxxttUUtt ==== ρρ
-- Hazard function
AHPHMADR IIIIIfor
VUyxgI
,,,,
)( ......... −=
**
*
tttdt
dt
tdt
d
∂•∂+
∂•∂=
∂•∂+
∂•∂=•
t* is the time since the last spike;
). 0.0117 0.072 0.257 1.1210(6.1exp=)( 4323 TTTTUA −−−−⋅ −
Application
A – solution in case of steady stimulation (self-similar);B – solution in case of abrupt excitation
0~
2~))((
~ 2
=
∂∂−−
∂∂+
∂∂
VVtU
Vtm
ρσρρτ
Single LIF neuron - Langevin equation
spikethenUVif T<
)()( ttUVdt
dVm ητ ++−=
0)( =>< tη)'()'()( 2 tttt m −>=< δτσηη
Fokker-Planck equation
0),(~ =TUtρ
0),(~ =−∞tρ
( )22)(exp1
),0(~ σσπ
ρ UVV −−=
σ))()((ˆ)( tUtVtu −=
σ))((ˆ)( tUUtT T −=
0~
2
1~~
=
∂∂−−
∂∂+
∂∂
uu
utm
ρρρτ0))(,(~ =tTtρ
0),(~ =−∞tρ
( )2exp1
),0(~ uu −=π
ρ
3. Hazard-function in the case of white noise-current(First-time passage problem)
BAH +≈Approximation:
)(
~
21
/)(~)(tTum
m utHtH
=∂∂−=≡ ρ
ττ
TUVm VtH
=∂∂−≡ ρ
τσ ~
2)(
2
Self-similar solution (T=const)
shapeutpamplitudet −− ),(,)(ρ
),()(),(~ utptut ρρ = ∫ ∞−=
)(),(~)(
tTduuttwhere ρρ
ptHup
puut
pm ⋅=
∂∂−−
∂∂+
∂∂
)(~21τ
Tuu
ptHwhere
=∂∂−=
21
)(~0),( =Ttp
0),( =−∞tp
Assumption.
( )2exp1
),0( uup −=π
U(t)=const (or T(t)=const). Notation:Then the shape of , which is , is invariable. ρ~ ),( utp
ptAu
ppu
u⋅=
∂∂−−
∂∂
)(2
1
Tuu
ptAwhere
=∂∂−=
2
1)(
0),( =Ttp
0),( =−∞tp
)(TAdt
dm ρρτ −=
),(~= tHdt
dm ρρτ −
HA ~≡
Equivalent formulation:
Frozen Gaussian distribution (dT/dt = ∞)
∫ ∞−=
)(),(~)(
tTduuttwhere ρρ
T(t) decreases fast.The initial Gaussian distribution remains almost unchanged except cutting at u=T.The hazard function in this case is H=B(T,dT/dt).
Assumption.
+
−=−=dt
dT
dT
d
dt
dB mm ρ
ρτρ
ρτ
For the simplicity, we consider the case of arbitrary but monotonically increasing T(t) and the Gaussian distribution
( )
<−=
otherwise
tTtuifuut
,0
)()(,exp1
),(~2
πρ
)(~
2 TFdt
dT
dt
dT
dT
dB m
m
++
−=
−= τρρ
τ
or
)erf(1)exp(2
)(~ 2
T
TTFwhere
+−=
π[x]+ for x>0 and zero otherwise
Bdt
dm ρρτ −=
U(t) UT
)(
~
2
1ˆ
tTum uH
=∂∂= ρ
τ
0~
2
1~~
=
∂∂−−
∂∂+
∂∂
uu
utmρρρτ
Approximation of hazard function in arbitrary case
BAH +≈
tt ∂∂= ρν )(
Weak stimulus Strong stimulus
0))(,(~ =tTtρ0),(~ =−∞tρ
( )2exp1
),0(~ uu −=π
ρ
Approximation:
σ))((ˆ)( tUUtTгдеT −=
A – solution in case of steady stimulation (self-similar);B – solution in case of abrupt excitation
Approximation of H by A is green, by B is blue, by A+B is red, exact solution is black.
Langevin equation
TUV <
),( tUIdt
dUC tot−=
σσησ
ξττ
τ
/))(,()(~,/)(,/))(,(
)(2
)(~),(),(
UUtUgtT
tqUVtUguгде
tqdtdq
tTutqudtdu
tU
Ttot
tot
m
−=
=−=
+−=
<+−=
)(2 tdtd ξστηητ +−=
0)( =>< tξ)'()'()( tttt −>=< δτξξ
0~1~~
~2
2
=∂∂−
−∂∂+
+−∂∂+
∂∂
qqu
ut m
ρτ
ρτ
ρτ
ρFokker-Planck eq.
0)~,~,(~),,(~),,(~),,(~
=≤==+∞===−∞==∞=
TqTutqut
qutqut
ρρρρ
[ ]
+−+−++== ququkkk
k
kqut 2)1(
21
exp21
),,0(~ 22
πρ
3. Hazard-function in the case of colored noise
)(),( ttVIdtdV
C tot η+−=
Without noise: TUU <
With noise:
or
or
shapeutpamplitudet −− ),(,)(ρ),,()(),,(~ qutptqut ρρ = 1),,(
)( =∫∫ ∞−
∞
∞−
tTduqutpdqwhere
ptHqp
qpq
kpquut
pm )(~)( 2
2
+
∂∂+
∂∂+−
∂∂=
∂∂τ
dqqTtpTqtUHwhereT
),~,( )~())((~~
−≡ ∫∞
. ),,(~=)(
~
dudqquttT
ρρ ∫∫∞
∞−∞−
ττ )/,(),( tUtUk m≡, ),~,(~ )~(
1))((~
~dqqTtTqtUH
T
ρρ
−≡ ∫∞
0))(~),(~,(),,(
),,(),,(
=≤==+∞===−∞==∞=
tTqtTutpqutp
qutpqutp
),(~= tHdt
dm ρρτ −
Self-similar solution (T=const)
Assumption. U(t) (or T(t)) is constant or slow. Then the shape of , which is , is invariable. ρ~ ),,( qutp
0)( 2
2
=+
∂∂+
∂∂+−
∂∂
pAqp
qpq
kpquu 0)~,~,(),,(
),,(),,(
=≤==+∞===−∞==∞=
TqTutpqutp
qutpqutp
dqqTtpTqAwhereT
),~,( )~(~
−= ∫∞
u
q
)T0.0117 -T0.072 -T0.257 - T1.12-exp(0.0061 (T)A 432=∞
21~ k
TT+=
Approximation of Hby A is green, by B is blue,by A+B is red,exact solution is black.
Hazard function in arbitrary case
tt ∂∂= ρν )(
K=1:
K=8:
BAH +≈Weak stimulus
Weak stimulus
Strong stimulus
Strong stimulus
Single cell level
Populations
Htt
ρρρ −=∂∂+
∂∂
∗
iAHPLHMADR IIIIIIIt
U
t
UC −−−−−−−=
∂∂+
∂∂
*
)()(
,)(
)(
*
*
U
yUy
t
y
t
y
U
xUx
t
x
t
x
y
x
τ
τ−=
∂∂+
∂∂
−=∂∂+
∂∂
∞
∞
)()0,(0
tdtFt νρρ ≡= ∫∞
+
∗
AHPHMADR IIIIIfor
VUyxgI
,,,,
)( ......... −=
t* is the time since the last spike
CBRD
Large-scale simulations(NMM & FR-models
for EEG & MRI)
Simulations by CBRD-model
Simulations. Current-step stimulation. Comparison with Monte-Carlo.
Non-adaptive neurons
(4000)
Simulations. Current-step stimulation. Color noise. Adaptive neurons.
LIF
Adaptive conductance-based neuron
with IM
Simulations. Oscillatory input.
Simulations. Constant current stimulation.Comparison with analytical solution.
[Johannesma 1968]
Simulations. Constant current stimulation.Color noise. Comparison with analytical solution.
Firing rate depends on the noise time constant.
1'
00 )/(
)(exp=
−
′′−
−− ∫∫ uduadu
uauHu ma
m
ττν
)(/= LT
La VUgIa −
dots – Monte-Carlosolid – eq.(*)dash – adiabatic approach [Moreno-Bote, Parga 2004]
(*)
GABA-IPSC AMPA-EPSCAMPA-EPSC
AMPA-EPSP
AMPA-EPSP
GABA-IPSP
GABA-IPSC
GABA-IPSPPSP
PSP
Firing rateFiring rate
SpikeSpike
Membraneequations
Threshold criterium
Population model
Synaptic current kinetics
Interconnected populations
Excitatory synaptic current:
- maximum specific conductance, - non-dimensional conductance
- reversal potential
Inhibitory synaptic current:
Non-dimensional synaptic conductances:
where
- rise and decay time constants- firing rate on j-type axonal terminals
NMDAGABAAMPAj
jjjd
jrj
jdj
rj Sm
dt
dm
dt
md
,,
2
2
),()(
=
=+++ νττττ
)()()( NMDANMDANMDANMDANMDA VVVftmgi −=t, ms
PS
C,p
A
0 10 20 30 40 50-50
0
50
100
150
200
Vh=-80 mV
Vh=+20 mV
Vh=-40 mV
AMPA-PSC(with PTX, APV)
experimentmodel
t, ms
PS
C,p
A
0 25 50 75 100-80
-60
-40
-20
0
20
40
Vh=+20 mV
Vh=-40 mV
NMDA-PSC(with PTX, CNQX)
Vh=+20 mV
Vh=-40 mV
NMDA-PSC(with PTX, CNQX)
Vh=+20 mV
Vh=-40 mV
NMDA-PSC(with PTX, CNQX)
experimentmodel
)()( AMPAAMPAAMPAAMPA
NMDAAMPAE
VVtmgi
iii
−=+=
))062.0exp(57.3/1/(1)( VMgVf NMDA −+=
jg
jm
jV
)()( GABAGABAGABAI VVtmgi −=
1))2exp(1(2)(S jj −ντ−+=νdj
rj ττ ,
sµτ 1=
)(tjν
t, ms
PS
C,p
A
0 10 20 30 40 50-500
-400
-300
-200
-100
0
Vh=-60 mV
fast GABA-A -IPSC(with CNQX, D,L-APV)
experimentmodel
t, ms
PS
C,p
A
0 10 20 30 40 50
-100
0
100
200
300
400
500
Vh=-80 mV
Vh=+20 mV
Vh=-40 mV
AMPA-PSC(with PTX, APV)
experimentmodel
t, ms
PS
C,p
A
0 25 50 75 100
-150
-100
-50
0
50
100
150 NMDA-PSC(with PTX, CNQX)NMDA-PSC(with PTX, CNQX)
Vh=+20 mV
Vh=-40 mV
NMDA-PSC(with PTX, CNQX)
experimentmodel
Pyramidal neurons
Interneurons
Approximations of synaptic currents
t, ms
PS
C,p
A
0 10 20 30 40 50
-100
-50
0
Vh=-64 mV
GABA-PSC(with CNQX, D-AP5)
experimentmodel
with IM and IAHP
),)(()()(
),()()(
SSS
Sexti
VtUtgtI
tItItI
−=+=
)(2
mS/cm 1
mV, 5
ms, 1
ms, 5.4
150
2
S
restatmV
g
V
pAI
V
S
S
ext
==
====
σ
ττ
)0,()()(
2)(
2
22 tgtg
dt
tdg
dt
tgdSS
SS
SS ρτττ =++
[S.Karnup, A.Stelzer 2001]
Experiment
Simulations. Interictal activity. Recurrent network of pyramidal cells,including all-to-all connectivity by excitatory synapses.
Model
Simulations. Gamma rhythm. Recurrent network of interneurons ,including all-to-all connectivity by inhibitory synapses
),)(()()(
),()()(
SSS
Sexti
VtUtgtI
tItItI
−=+=
)0,()(
)()()(
2)(
*
2
22
==
−=++
tttapproachdensityfor
tgtgdt
tdg
dt
tgddSS
SS
SS
ρν
τντττ
2
S
d
S
7mS/cm
-80mV,V
1ms,
1ms,
3ms,
=
====
Sg
τττ
OscillationsModel ExperimentsControl (“Kainate”) +“Bicuculline”
Spikes in single neurons
Conductances
Power Spectrum of Extracellular Potentials
Spike timing of pyramidal and inhibitory cells.
[Khazipov, Holmes, 2003]Kainate-induced oscillations in CA3.
[A.Fisahn et al., 1998] Cholinergically induced oscillations in CA3
[N.Hajos, J.Palhalmi, E.O.Mann, B.Nemeth, O.Paulsen, and T.F.Freund. J.Neuroscience, 24(41):9127–9137, 2004]
conbic
All the simulations were done with a single set of parameters. All the parameters except synaptic maximum conductances have been obtained by fitting to experimental registration of elementary events such as patch-electrode current-induced traces, spike trains and monosynaptic responses .
The model reproduces the following characteristics of gamma-oscillations :
� frequency of population spikes
� a single pyramidal cell does not fire every cycle
� every interneuron fires every cycle
� amplitude of EPSC is less than that of IPSC
� blockage of GABA-A receptors reduces the frequency
� peak of pyramidal cell’s firing frequency corresponds to the descending phase of EPSC and the ascending phase of IPSC
� firing of interneurons follows the firing of pyramidal cells
� gamma-oscillations are homogeneous in space along the cortical surface (data not shown)
Spatial connections
22 )()(),,,( YyXxYXyxd −+−=
- firing rate on presynaptic terminals;- firing rate on somas.
Assumption: distances from soma to synapses have exponentially decreasingdistribution p(x) [B.Hellwig 2000].
[V.Jirsa, G.Haken 1996][P.Nunez 1995] [J.Wright, P.Robinson 1995]
),,(2 22
2
2
222
2
2
yxttyx
ctt
νγγφφφγφγφ
∂∂+=
∂∂+
∂∂−+
∂∂+
∂∂
),,( yxtφ),,( yxtν
where γ = c/λ; c – the average velocity of spike propagation along the cortex surface by axons; λ – characteristic axon length. [D.Contreras, R.Llinas 2001]
Experiment:
, ),,,(),,/),,,((=),,( dYdXYXyxWYXcYXyxdtyxt iij −∫∫ νϕ
λ),,,(
),,,(YXyxd
eYXyxW−
=
PSPs and PSCs evoked by extracellular stimulation and registered
at 3.5cm away, w/ and w/o kainate.
[S.Karnup, A.Stelzer1999] Effects of GABA-A receptor blockade on orthodromic potentials in CA1 pyramidal cells. Superimposed responses in a pyramidal cell soma before and after application of picrotoxin (PTX, 100 muM). Control and PTX recordings were obtained at V rest (-64 mV; 150 muAstimulation intensities; 1 mm distance between stratum radiatum stimulation site and perpendicular line through stratum pyramidale recording site). The recordings were carried out in ‘minislices’ in which the CA3 region was cut off by dissection.
[V.Crepel, R.Khazipov, Y.Ben-Ari, 1997] In normal concentrations of Mg and in the absence of CNQX, block of GABA-A receptors induced a late synaptic response.
BA
C
[B.Mlinar,A.M.Pugliese, R.Corradetti2001] Components of complex synaptic responses evoked in CA1 pyramidal neurones in the presence of GABAA receptor block.
The model reproduces postsynaptic currents and postsynaptic potentials registered on somas of pyramidal cells, namely:
� monosynaptic EPSCs and EPSPs
� disynaptic IPSC/Ps followed be EPSC/Ps
� polysynaptic EPSC/Ps
� reduction of delays in polysynaptic EPSCs
� decay of excitation after II component of poly-EPSCs in presence of GABA-A receptor block.
The model predicts that the evoked responses are essentially non-homogeneous in space:
Spatial profiles of membrane potential and firing rate in pyramids.
Evoked responsesModel Experiments
WavesIn the case of reduced GABA-reversal potential VGABA= -50mV and stimulation by extracellular electrode we obtain a traveling wave of stable amplitude and velocity 0.15 m/s. The velocity is much less than the axon propagation velocity (1m/s) and is determined mostly by synaptic interactions.
ms
Hz mV
0 25 50 75 1000
20
40
60
80
100
120
140 voltage, pyramidsvoltage, interneuronsfiring rate, pyramidsfiring rate, interneurons
-40
-60
B
mm
Hz mV
10 20 30 400
20
40
60
80
100
120 voltage, pyramidsvoltage, interneuronsfiring rate, pyramidsfiring rate, interneurons
0.15m/s -40
-60
Fig.5. Wave propagating from the site of extracellular stimulation at right border of the “slice”.A, Evoked responses of pyramidal cells and interneurons at the site of stimulation. B, Profiles of mean voltage and firing rate in pyramidal cells and interneurons at the time 200 ms after the stimulus.
A
[Leinekugel et al. 1998]. Spontaneous GDPs propagate synchronously in both hippocampi from septal to temporal poles. Multiple extracellular field recordings from the CA3 region of the intact bilateral septohippocampal complex. Simultaneous extracellular field recordings at the four recording sites indicated in the scheme. Corresponding electrophysiological traces (1–4) showing propagation of a GDP at a large time scale.
[D.Golomb, Y.Amitai, 1997]Propagation of discharges in disinhibitedneocortical slices.
Model Experiments
Waves with unchanging chape and velocity are observed in cortical tissue in disinhibiting or overexciting conditions. The waves are produced by complex interaction of pyramidal cells and interneurons. That is confirmed by much lower speed of the wave propagation comparing with the axon propagation velocity which is the coefficient in the wave-like equation.Analysis of wave solutions and more detailed comparison with experiments are expected in future.
From CBRD to Firing-Rate model
Macro- and meso-scale
internal granularlayer
internal pyramidallayer
external pyramidallayer
external granularlayer
AP generation zone synapses
macro-scale meso-scale micro-scale
[S.Kiebel][C.Friston]
IVUggdt
dUC LSL −−+−= ))((
( )
)(;)(
exp1
-)(
(steady) ;)(1)(
/
),()()(
2
2
m
1/)(
/)(
2
suddenUV
dt
dUUB
duuerfeUA
gC
dtdUUBUAt
T
UV
UV
um
Lm
T
reset
−−
=
+=
=+=
+
−−
−∫
σσπτ
πτ
τν
σ
σ
Hazard-function:-- firing rate
Oscillating input
Firing-rate model
[Chizhov, Rodrigues, Terry // Phys.Lett.A, 2007 ]
( )
)(;)(
exp1
-)(
(steady) ;)(1)(
/
),()()(
2
2
m
1/)(
/)(
2
suddenUV
dt
dUUB
duuerfeUA
gC
dtdUUBUAt
T
UV
UV
um
Lm
T
reset
−−
=
+=
=+=
+
−−
−∫
σσπτ
πτ
τν
σ
σ
Hazard-function:-- firing rate
Oscillating input
[Чижов, Бучин // Нейроинформатика-2009 ]
IVUtwgVUtngVUggdt
dUC AHPAHPMMLSL +−−−−−+−= ))(())(())(( 2
)()/1,/1(
)1()(
0101
2
201 tv
K
www
dt
dw
dt
wd
AHPAHPAHPAHPAHPAHP ττ
χττττ −=+−++ ∞
)()/1,/1(
)1()(
0101
2
201 tv
K
nnn
dt
dn
dt
nd
MMMMMM ττ
ξττττ −=+−++ ∞
Not-adaptive neurons Adaptive neurons
Рис. 12. Схема активности
популяции FS (fast spiking) нейронов, возбуждаемых
внешним стимулом νext(t), приходящим из таламуса. Обозначения: ν(t) –популяционная частота спайков
FS нейронов, gE(t), gI(t) –проводимости возбуждающих и
тормозящих синапсов.
FSν
ext
ν gI
gE
Experiment
Model
Рис. 13. Постсинаптический
(моносинаптический) ток в FS-нейроне при слабой
таламической стимуляции
током 30 µA и потенциале
фиксации -88 mV в
эксперименте (вверху) (adaptedby permission from MacmillanPublishers Ltd: (Cruikshank et al., 2007), copyright 2007) и в модели
(внизу).
Рис. 14. Ответы FS-нейронов на таламическую стимуляцию
током 120 µA в эксперименте (слева) (adapted by permission fromMacmillan Publishers Ltd: (Cruikshank et al., 2007), © 2007) и в
модели (справа). A, B - постсинаптические токи при
потенциале фиксации -88, -62, и -35 mV; C, D - синаптические
проводимости; E, F – постсинаптические потенциалы U и
модельная популяционная частота ν.
))(()()( EEE VtVtgti −=
)()( 212
2
21 tggdt
dgdt
gd extEE
EEEEEE ντττττ =+++
),()()( titiVUgdt
dUC IELL ++−−=
),()()( UBUAt +=ν
( ) ;)(1)(
-12/)(
2/)(
2
+= ∫
−
−
VT
Vreset
UV
UV
um duuerfeUA
σ
σπτ
−−×
=+
2
2
2)(
exp21
)(V
T
V
UV
dt
dUUB
σσπ
))(()()( III VtVtgti −=
)()( 212
2
21 tggdt
dg
dt
gdII
IIIIII ντττττ =+++
Simple model of interacting cortical interneurons, evoked by thalamus
Синаптические токи и проводимости:
Мембранный потенциал:
Популяционная частота спайков:
Частотная модель популяции
адаптивных нейронов: «интериктальная» активность
EI
)(),( νν MAHP II
)(νSI
)/,()()(
)()()(
2)(
))((
)()()()(
:
2
22
dtdUUBUAt
tvgtgdt
tgd
dt
tgd
VVtgI
IVVgIIIt
VC
модельFR
SSS
SS
S
SSS
SLLMAHP
+=
=++
−=
−−−−−=∂∂
ν
τττ
ννν
• What can be modeled on population level?
• Which details are important?
• What kinds of population models do exist?
• Which one to choose?
Monte-Carlo simulations:
conventionalFiring-Rate model:
CBRD:
N
ttn
tt
спайкиVVVV
tVVggIt
VC
КарлоМонтеМетод
act
resetT
ILSL
)(1)(
т , если
)())((
:
∆+∆
=
=>
+−+−=∂∂
−
ν
ξσ
∫∞
+
==
+=
−+−=
∂∂+
∂∂
−=∂∂+
∂∂
0
*)0,()(
))/,()((1
*)),((
))((*
*
:
dtHttv
dtdUUBUAttUH
VUggIt
U
t
UC
Htt
модельRD
m
LSL
ρρ
τ
ρρρ
)/,()()(
))((
:FR
dtdUUBUAt
VUggIdt
dUC
модель
LSL
+=
−+−=
ν
Mathematical complexity:104 ODEs 1 ODE a few ODEs 1-d PDEs
Precision:4 2 3 5
Precision for non-stationary problems:5 2 4 5
Precision for adaptive neurons :5 1 3 4
Computational efficiency:2 5 5 4
Mathematical analyzability:1 5 4 4
)/,()()(
)()()(
2)(
)()())((
:
2
22
dtdUUBUAt
tvgtgdt
tgd
dt
tgd
IIVVggItV
C
модельFR
SSS
SS
S
AHPMLSL
+=
=++
−−−+−=∂∂
ν
τττ
νν
modified Firing-Rate model (non-stationary and adaptive):