| 1 |
|
|
| 2 |
|
function MAP1_14AP(inputSignal, sampleRate, BFlist, MAPparamsName, ...
|
| 3 |
|
AN_spikesOrProbability, paramChanges)
|
| 4 |
|
% To test this function use test_MAP1_14 in this folder
|
| 5 |
|
%
|
| 6 |
|
% All arguments are mandatory.
|
| 7 |
|
%
|
| 8 |
|
% BFlist is a vector of BFs but can be '-1' to allow MAPparams to choose
|
| 9 |
|
% MAPparamsName='Normal'; % source of model parameters
|
| 10 |
|
% AN_spikesOrProbability='spikes'; % or 'probability'
|
| 11 |
|
% paramChanges is a cell array of strings that can be used to make last
|
| 12 |
|
% minute parameter changes, e.g., to simulate OHC loss
|
| 13 |
|
% e.g. paramChanges{1}= 'DRNLParams.a=0;'; % disable OHCs
|
| 14 |
|
% e.g. paramchanges={}; % no changes
|
| 15 |
|
% The model parameters are established in the MAPparams<***> file
|
| 16 |
|
% and stored as global
|
| 17 |
|
|
| 18 |
|
restorePath=path;
|
| 19 |
|
addpath (['..' filesep 'parameterStore'])
|
| 20 |
|
|
| 21 |
|
|
| 22 |
|
CONVOLUTION_CHANGE_TEST = 0; %for debug
|
| 23 |
|
|
| 24 |
|
|
| 25 |
|
global OMEParams DRNLParams IHC_cilia_RPParams IHCpreSynapseParams
|
| 26 |
|
global AN_IHCsynapseParams MacGregorParams MacGregorMultiParams
|
| 27 |
|
|
| 28 |
|
% All of the results of this function are stored as global
|
| 29 |
|
global dt ANdt savedBFlist saveAN_spikesOrProbability saveMAPparamsName...
|
| 30 |
|
savedInputSignal OMEextEarPressure TMoutput OMEoutput ARattenuation ...
|
| 31 |
|
DRNLoutput IHC_cilia_output IHCrestingCiliaCond IHCrestingV...
|
| 32 |
|
IHCoutput ANprobRateOutput ANoutput savePavailable tauCas ...
|
| 33 |
|
CNoutput ICoutput ICmembraneOutput ICfiberTypeRates MOCattenuation
|
| 34 |
|
|
| 35 |
|
% Normally only ICoutput(logical spike matrix) or ANprobRateOutput will be
|
| 36 |
|
% needed by the user; so the following will suffice
|
| 37 |
|
% global ANdt ICoutput ANprobRateOutput
|
| 38 |
|
|
| 39 |
|
% Note that sampleRate has not changed from the original function call and
|
| 40 |
|
% ANprobRateOutput is sampled at this rate
|
| 41 |
|
% However ANoutput, CNoutput and IC output are stored as logical
|
| 42 |
|
% 'spike' matrices using a lower sample rate (see ANdt).
|
| 43 |
|
|
| 44 |
|
% When AN_spikesOrProbability is set to probability,
|
| 45 |
|
% no spike matrices are computed.
|
| 46 |
|
% When AN_spikesOrProbability is set to 'spikes',
|
| 47 |
|
% no probability output is computed
|
| 48 |
|
|
| 49 |
|
% Efferent control variables are ARattenuation and MOCattenuation
|
| 50 |
|
% These are scalars between 1 (no attenuation) and 0.
|
| 51 |
|
% They are represented with dt=1/sampleRate (not ANdt)
|
| 52 |
|
% They are computed using either AN probability rate output
|
| 53 |
|
% or IC (spikes) output as approrpriate.
|
| 54 |
|
% AR is computed using across channel activity
|
| 55 |
|
% MOC is computed on a within-channel basis.
|
| 56 |
|
|
| 57 |
|
if nargin<1
|
| 58 |
|
error(' MAP1_14 is not a script but a function that must be called')
|
| 59 |
|
end
|
| 60 |
|
|
| 61 |
|
if nargin<6
|
| 62 |
|
paramChanges=[];
|
| 63 |
|
end
|
| 64 |
|
% Read parameters from MAPparams<***> file in 'parameterStore' folder
|
| 65 |
|
% Beware, 'BFlist=-1' is a legitimate argument for MAPparams<>
|
| 66 |
|
% It means that the calling program allows MAPparams to specify the list
|
| 67 |
|
cmd=['method=MAPparams' MAPparamsName ...
|
| 68 |
|
'(BFlist, sampleRate, 0, paramChanges);'];
|
| 69 |
|
eval(cmd);
|
| 70 |
|
BFlist=DRNLParams.nonlinCFs;
|
| 71 |
|
|
| 72 |
|
% save as global for later plotting if required
|
| 73 |
|
savedBFlist=BFlist;
|
| 74 |
|
saveAN_spikesOrProbability=AN_spikesOrProbability;
|
| 75 |
|
saveMAPparamsName=MAPparamsName;
|
| 76 |
|
|
| 77 |
|
dt=1/sampleRate;
|
| 78 |
|
duration=length(inputSignal)/sampleRate;
|
| 79 |
|
% segmentDuration is specified in parameter file (must be >efferent delay)
|
| 80 |
|
segmentDuration=method.segmentDuration;
|
| 81 |
|
segmentLength=round(segmentDuration/ dt);
|
| 82 |
|
segmentTime=dt*(1:segmentLength); % used in debugging plots
|
| 83 |
|
|
| 84 |
|
% all spiking activity is computed using longer epochs
|
| 85 |
|
ANspeedUpFactor=AN_IHCsynapseParams.ANspeedUpFactor; % e.g.5 times
|
| 86 |
|
|
| 87 |
|
% inputSignal must be row vector
|
| 88 |
|
[r c]=size(inputSignal);
|
| 89 |
|
if r>c, inputSignal=inputSignal'; end % transpose
|
| 90 |
|
% ignore stereo signals
|
| 91 |
|
inputSignal=inputSignal(1,:); % drop any second channel
|
| 92 |
|
savedInputSignal=inputSignal;
|
| 93 |
|
|
| 94 |
|
% Segment the signal
|
| 95 |
|
% The sgment length is given but the signal length must be adjusted to be a
|
| 96 |
|
% multiple of both the segment length and the reduced segmentlength
|
| 97 |
|
[nSignalRows signalLength]=size(inputSignal);
|
| 98 |
|
segmentLength=ceil(segmentLength/ANspeedUpFactor)*ANspeedUpFactor;
|
| 99 |
|
% Make the signal length a whole multiple of the segment length
|
| 100 |
|
nSignalSegments=ceil(signalLength/segmentLength);
|
| 101 |
|
padSize=nSignalSegments*segmentLength-signalLength;
|
| 102 |
|
pad=zeros(nSignalRows,padSize);
|
| 103 |
|
inputSignal=[inputSignal pad];
|
| 104 |
|
[ignore signalLength]=size(inputSignal);
|
| 105 |
|
|
| 106 |
|
% AN (spikes) is computed at a lower sample rate when spikes required
|
| 107 |
|
% so it has a reduced segment length (see 'ANspeeUpFactor' above)
|
| 108 |
|
% AN CN and IC all use this sample interval
|
| 109 |
|
ANdt=dt*ANspeedUpFactor;
|
| 110 |
|
reducedSegmentLength=round(segmentLength/ANspeedUpFactor);
|
| 111 |
|
reducedSignalLength= round(signalLength/ANspeedUpFactor);
|
| 112 |
|
|
| 113 |
|
%% Initialise with respect to each stage before computing
|
| 114 |
|
% by allocating memory,
|
| 115 |
|
% by computing constants
|
| 116 |
|
% by establishing easy to read variable names
|
| 117 |
|
% The computations are made in segments and boundary conditions must
|
| 118 |
|
% be established and stored. These are found in variables with
|
| 119 |
|
% 'boundary' or 'bndry' in the name
|
| 120 |
|
|
| 121 |
|
%% OME ---
|
| 122 |
|
% external ear resonances
|
| 123 |
|
OMEexternalResonanceFilters=OMEParams.externalResonanceFilters;
|
| 124 |
|
[nOMEExtFilters c]=size(OMEexternalResonanceFilters);
|
| 125 |
|
% details of external (outer ear) resonances
|
| 126 |
|
OMEgaindBs=OMEexternalResonanceFilters(:,1);
|
| 127 |
|
OMEgainScalars=10.^(OMEgaindBs/20);
|
| 128 |
|
OMEfilterOrder=OMEexternalResonanceFilters(:,2);
|
| 129 |
|
OMElowerCutOff=OMEexternalResonanceFilters(:,3);
|
| 130 |
|
OMEupperCutOff=OMEexternalResonanceFilters(:,4);
|
| 131 |
|
% external resonance coefficients
|
| 132 |
|
ExtFilter_b=cell(nOMEExtFilters,1);
|
| 133 |
|
ExtFilter_a=cell(nOMEExtFilters,1);
|
| 134 |
|
for idx=1:nOMEExtFilters
|
| 135 |
|
Nyquist=sampleRate/2;
|
| 136 |
|
[b, a] = butter(OMEfilterOrder(idx), ...
|
| 137 |
|
[OMElowerCutOff(idx) OMEupperCutOff(idx)]...
|
| 138 |
|
/Nyquist);
|
| 139 |
|
ExtFilter_b{idx}=b;
|
| 140 |
|
ExtFilter_a{idx}=a;
|
| 141 |
|
end
|
| 142 |
|
OMEExtFilterBndry=cell(2,1);
|
| 143 |
|
OMEextEarPressure=zeros(1,signalLength); % pressure at tympanic membrane
|
| 144 |
|
|
| 145 |
|
% pressure to velocity conversion using smoothing filter (50 Hz cutoff)
|
| 146 |
|
tau=1/(2*pi*50);
|
| 147 |
|
a1=dt/tau-1; a0=1;
|
| 148 |
|
b0=1+ a1;
|
| 149 |
|
TMdisp_b=b0; TMdisp_a=[a0 a1];
|
| 150 |
|
% figure(9), freqz(TMdisp_b, TMdisp_a)
|
| 151 |
|
OME_TMdisplacementBndry=[];
|
| 152 |
|
|
| 153 |
|
% OME high pass (simulates poor low frequency stapes response)
|
| 154 |
|
OMEhighPassHighCutOff=OMEParams.OMEstapesLPcutoff;
|
| 155 |
|
Nyquist=sampleRate/2;
|
| 156 |
|
[stapesDisp_b,stapesDisp_a] = butter(1, OMEhighPassHighCutOff/Nyquist, 'high');
|
| 157 |
|
% figure(10), freqz(stapesDisp_b, stapesDisp_a)
|
| 158 |
|
|
| 159 |
|
OMEhighPassBndry=[];
|
| 160 |
|
|
| 161 |
|
% OMEampStapes might be reducdant (use OMEParams.stapesScalar)
|
| 162 |
|
stapesScalar= OMEParams.stapesScalar;
|
| 163 |
|
|
| 164 |
|
% Acoustic reflex
|
| 165 |
|
efferentDelayPts=round(OMEParams.ARdelay/dt);
|
| 166 |
|
% smoothing filter
|
| 167 |
|
a1=dt/OMEParams.ARtau-1; a0=1;
|
| 168 |
|
b0=1+ a1;
|
| 169 |
|
ARfilt_b=b0; ARfilt_a=[a0 a1];
|
| 170 |
|
|
| 171 |
|
ARattenuation=ones(1,signalLength);
|
| 172 |
|
ARrateThreshold=OMEParams.ARrateThreshold; % may not be used
|
| 173 |
|
ARrateToAttenuationFactor=OMEParams.rateToAttenuationFactor;
|
| 174 |
|
ARrateToAttenuationFactorProb=OMEParams.rateToAttenuationFactorProb;
|
| 175 |
|
ARboundary=[];
|
| 176 |
|
ARboundaryProb=0;
|
| 177 |
|
|
| 178 |
|
% save complete OME record (stapes displacement)
|
| 179 |
|
OMEoutput=zeros(1,signalLength);
|
| 180 |
|
TMoutput=zeros(1,signalLength);
|
| 181 |
|
|
| 182 |
|
%% BM ---
|
| 183 |
|
% BM is represented as a list of locations identified by BF
|
| 184 |
|
DRNL_BFs=BFlist;
|
| 185 |
|
nBFs= length(DRNL_BFs);
|
| 186 |
|
|
| 187 |
|
% DRNLchannelParameters=DRNLParams.channelParameters;
|
| 188 |
|
DRNLresponse= zeros(nBFs, segmentLength);
|
| 189 |
|
|
| 190 |
|
MOCrateToAttenuationFactor=DRNLParams.rateToAttenuationFactor;
|
| 191 |
|
rateToAttenuationFactorProb=DRNLParams.rateToAttenuationFactorProb;
|
| 192 |
|
MOCrateThresholdProb=DRNLParams.MOCrateThresholdProb;
|
| 193 |
|
|
| 194 |
|
% smoothing filter for MOC
|
| 195 |
|
a1=dt/DRNLParams.MOCtau-1; a0=1;
|
| 196 |
|
b0=1+ a1;
|
| 197 |
|
MOCfilt_b=b0; MOCfilt_a=[a0 a1];
|
| 198 |
|
% figure(9), freqz(stapesDisp_b, stapesDisp_a)
|
| 199 |
|
MOCboundary=cell(nBFs,1);
|
| 200 |
|
MOCprobBoundary=cell(nBFs,1);
|
| 201 |
|
|
| 202 |
|
MOCattSegment=zeros(nBFs,reducedSegmentLength);
|
| 203 |
|
MOCattenuation=ones(nBFs,signalLength);
|
| 204 |
|
|
| 205 |
|
if DRNLParams.a>0
|
| 206 |
|
DRNLcompressionThreshold=10^((1/(1-DRNLParams.c))* ...
|
| 207 |
|
log10(DRNLParams.b/DRNLParams.a));
|
| 208 |
|
else
|
| 209 |
|
DRNLcompressionThreshold=inf;
|
| 210 |
|
end
|
| 211 |
|
|
| 212 |
|
DRNLlinearOrder= DRNLParams.linOrder;
|
| 213 |
|
DRNLnonlinearOrder= DRNLParams.nonlinOrder;
|
| 214 |
|
|
| 215 |
|
DRNLa=DRNLParams.a;
|
| 216 |
|
DRNLb=DRNLParams.b;
|
| 217 |
|
DRNLc=DRNLParams.c;
|
| 218 |
|
linGAIN=DRNLParams.g;
|
| 219 |
|
%
|
| 220 |
|
% gammatone filter coefficients for linear pathway
|
| 221 |
|
bw=DRNLParams.linBWs';
|
| 222 |
|
phi = 2 * pi * bw * dt;
|
| 223 |
|
cf=DRNLParams.linCFs';
|
| 224 |
|
theta = 2 * pi * cf * dt;
|
| 225 |
|
cos_theta = cos(theta);
|
| 226 |
|
sin_theta = sin(theta);
|
| 227 |
|
alpha = -exp(-phi).* cos_theta;
|
| 228 |
|
b0 = ones(nBFs,1);
|
| 229 |
|
b1 = 2 * alpha;
|
| 230 |
|
b2 = exp(-2 * phi);
|
| 231 |
|
z1 = (1 + alpha .* cos_theta) - (alpha .* sin_theta) * i;
|
| 232 |
|
z2 = (1 + b1 .* cos_theta) - (b1 .* sin_theta) * i;
|
| 233 |
|
z3 = (b2 .* cos(2 * theta)) - (b2 .* sin(2 * theta)) * i;
|
| 234 |
|
tf = (z2 + z3) ./ z1;
|
| 235 |
|
a0 = abs(tf);
|
| 236 |
|
a1 = alpha .* a0;
|
| 237 |
|
GTlin_a = [b0, b1, b2];
|
| 238 |
|
GTlin_b = [a0, a1];
|
| 239 |
|
GTlinOrder=DRNLlinearOrder;
|
| 240 |
|
GTlinBdry=cell(nBFs,GTlinOrder);
|
| 241 |
|
|
| 242 |
|
% nonlinear gammatone filter coefficients
|
| 243 |
|
bw=DRNLParams.nlBWs';
|
| 244 |
|
phi = 2 * pi * bw * dt;
|
| 245 |
|
cf=DRNLParams.nonlinCFs';
|
| 246 |
|
theta = 2 * pi * cf * dt;
|
| 247 |
|
cos_theta = cos(theta);
|
| 248 |
|
sin_theta = sin(theta);
|
| 249 |
|
alpha = -exp(-phi).* cos_theta;
|
| 250 |
|
b0 = ones(nBFs,1);
|
| 251 |
|
b1 = 2 * alpha;
|
| 252 |
|
b2 = exp(-2 * phi);
|
| 253 |
|
z1 = (1 + alpha .* cos_theta) - (alpha .* sin_theta) * i;
|
| 254 |
|
z2 = (1 + b1 .* cos_theta) - (b1 .* sin_theta) * i;
|
| 255 |
|
z3 = (b2 .* cos(2 * theta)) - (b2 .* sin(2 * theta)) * i;
|
| 256 |
|
tf = (z2 + z3) ./ z1;
|
| 257 |
|
a0 = abs(tf);
|
| 258 |
|
a1 = alpha .* a0;
|
| 259 |
|
GTnonlin_a = [b0, b1, b2];
|
| 260 |
|
GTnonlin_b = [a0, a1];
|
| 261 |
|
GTnonlinOrder=DRNLnonlinearOrder;
|
| 262 |
|
GTnonlinBdry1=cell(nBFs, GTnonlinOrder);
|
| 263 |
|
GTnonlinBdry2=cell(nBFs, GTnonlinOrder);
|
| 264 |
|
|
| 265 |
|
% complete BM record (BM displacement)
|
| 266 |
|
DRNLoutput=zeros(nBFs, signalLength);
|
| 267 |
|
|
| 268 |
|
|
| 269 |
|
%% IHC ---
|
| 270 |
|
% IHC cilia activity and receptor potential
|
| 271 |
|
% viscous coupling between BM and stereocilia displacement
|
| 272 |
|
% Nyquist=sampleRate/2;
|
| 273 |
|
% IHCcutoff=1/(2*pi*IHC_cilia_RPParams.tc);
|
| 274 |
|
% [IHCciliaFilter_b,IHCciliaFilter_a]=...
|
| 275 |
|
% butter(1, IHCcutoff/Nyquist, 'high');
|
| 276 |
|
a1=dt/IHC_cilia_RPParams.tc-1; a0=1;
|
| 277 |
|
b0=1+ a1;
|
| 278 |
|
% high pass (i.e. low pass reversed)
|
| 279 |
|
IHCciliaFilter_b=[a0 a1]; IHCciliaFilter_a=b0;
|
| 280 |
|
% figure(9), freqz(IHCciliaFilter_b, IHCciliaFilter_a)
|
| 281 |
|
|
| 282 |
|
IHCciliaBndry=cell(nBFs,1);
|
| 283 |
|
|
| 284 |
|
% IHC apical conductance (Boltzman function)
|
| 285 |
|
IHC_C= IHC_cilia_RPParams.C;
|
| 286 |
|
IHCu0= IHC_cilia_RPParams.u0;
|
| 287 |
|
IHCu1= IHC_cilia_RPParams.u1;
|
| 288 |
|
IHCs0= IHC_cilia_RPParams.s0;
|
| 289 |
|
IHCs1= IHC_cilia_RPParams.s1;
|
| 290 |
|
IHCGmax= IHC_cilia_RPParams.Gmax;
|
| 291 |
|
IHCGa= IHC_cilia_RPParams.Ga; % (leakage)
|
| 292 |
|
|
| 293 |
|
IHCGu0 = IHCGa+IHCGmax./(1+exp(IHCu0/IHCs0).*(1+exp(IHCu1/IHCs1)));
|
| 294 |
|
IHCrestingCiliaCond=IHCGu0;
|
| 295 |
|
|
| 296 |
|
% Receptor potential
|
| 297 |
|
IHC_Cab= IHC_cilia_RPParams.Cab;
|
| 298 |
|
IHC_Gk= IHC_cilia_RPParams.Gk;
|
| 299 |
|
IHC_Et= IHC_cilia_RPParams.Et;
|
| 300 |
|
IHC_Ek= IHC_cilia_RPParams.Ek;
|
| 301 |
|
IHC_Ekp= IHC_Ek+IHC_Et*IHC_cilia_RPParams.Rpc;
|
| 302 |
|
|
| 303 |
|
IHCrestingV= (IHC_Gk*IHC_Ekp+IHCGu0*IHC_Et)/(IHCGu0+IHC_Gk);
|
| 304 |
|
|
| 305 |
|
IHC_Vnow= IHCrestingV*ones(nBFs,1); % initial voltage
|
| 306 |
|
IHC_RP= zeros(nBFs,segmentLength);
|
| 307 |
|
|
| 308 |
|
% complete record of IHC receptor potential (V)
|
| 309 |
|
IHCciliaDisplacement= zeros(nBFs,segmentLength);
|
| 310 |
|
IHCoutput= zeros(nBFs,signalLength);
|
| 311 |
|
IHC_cilia_output= zeros(nBFs,signalLength);
|
| 312 |
|
|
| 313 |
|
|
| 314 |
|
%% pre-synapse ---
|
| 315 |
|
% Each BF is replicated using a different fiber type to make a 'channel'
|
| 316 |
|
% The number of channels is nBFs x nANfiberTypes
|
| 317 |
|
% Fiber types are specified in terms of tauCa
|
| 318 |
|
nANfiberTypes= length(IHCpreSynapseParams.tauCa);
|
| 319 |
|
tauCas= IHCpreSynapseParams.tauCa;
|
| 320 |
|
nANchannels= nANfiberTypes*nBFs;
|
| 321 |
|
synapticCa= zeros(nANchannels,segmentLength);
|
| 322 |
|
|
| 323 |
|
% Calcium control (more calcium, greater release rate)
|
| 324 |
|
ECa=IHCpreSynapseParams.ECa;
|
| 325 |
|
gamma=IHCpreSynapseParams.gamma;
|
| 326 |
|
beta=IHCpreSynapseParams.beta;
|
| 327 |
|
tauM=IHCpreSynapseParams.tauM;
|
| 328 |
|
mICa=zeros(nANchannels,segmentLength);
|
| 329 |
|
GmaxCa=IHCpreSynapseParams.GmaxCa;
|
| 330 |
|
synapse_z= IHCpreSynapseParams.z;
|
| 331 |
|
synapse_power=IHCpreSynapseParams.power;
|
| 332 |
|
|
| 333 |
|
% tauCa vector is established across channels to allow vectorization
|
| 334 |
|
% (one tauCa per channel). Do not confuse with tauCas (one pre fiber type)
|
| 335 |
|
tauCa=repmat(tauCas, nBFs,1);
|
| 336 |
|
tauCa=reshape(tauCa, nANchannels, 1);
|
| 337 |
|
|
| 338 |
|
% presynapse startup values (vectors, length:nANchannels)
|
| 339 |
|
% proportion (0 - 1) of Ca channels open at IHCrestingV
|
| 340 |
|
mICaCurrent=((1+beta^-1 * exp(-gamma*IHCrestingV))^-1)...
|
| 341 |
|
*ones(nBFs*nANfiberTypes,1);
|
| 342 |
|
% corresponding startup currents
|
| 343 |
|
ICaCurrent= (GmaxCa*mICaCurrent.^3) * (IHCrestingV-ECa);
|
| 344 |
|
CaCurrent= ICaCurrent.*tauCa;
|
| 345 |
|
|
| 346 |
|
% vesicle release rate at startup (one per channel)
|
| 347 |
|
% kt0 is used only at initialisation
|
| 348 |
|
kt0= -synapse_z * CaCurrent.^synapse_power;
|
| 349 |
|
|
| 350 |
|
|
| 351 |
|
%% AN ---
|
| 352 |
|
% each row of the AN matrices represents one AN fiber
|
| 353 |
|
% The results computed either for probabiities *or* for spikes (not both)
|
| 354 |
|
% Spikes are necessary if CN and IC are to be computed
|
| 355 |
|
nFibersPerChannel= AN_IHCsynapseParams.numFibers;
|
| 356 |
|
nANfibers= nANchannels*nFibersPerChannel;
|
| 357 |
|
AN_refractory_period= AN_IHCsynapseParams.refractory_period;
|
| 358 |
|
|
| 359 |
|
y=AN_IHCsynapseParams.y;
|
| 360 |
|
l=AN_IHCsynapseParams.l;
|
| 361 |
|
x=AN_IHCsynapseParams.x;
|
| 362 |
|
r=AN_IHCsynapseParams.r;
|
| 363 |
|
M=round(AN_IHCsynapseParams.M);
|
| 364 |
|
|
| 365 |
|
% probability (NB initial 'P' on everything)
|
| 366 |
|
PAN_ydt = repmat(AN_IHCsynapseParams.y*dt, nANchannels,1);
|
| 367 |
|
PAN_ldt = repmat(AN_IHCsynapseParams.l*dt, nANchannels,1);
|
| 368 |
|
PAN_xdt = repmat(AN_IHCsynapseParams.x*dt, nANchannels,1);
|
| 369 |
|
PAN_rdt = repmat(AN_IHCsynapseParams.r*dt, nANchannels,1);
|
| 370 |
|
PAN_rdt_plus_ldt = PAN_rdt + PAN_ldt;
|
| 371 |
|
PAN_M=round(AN_IHCsynapseParams.M);
|
| 372 |
|
|
| 373 |
|
% compute starting values
|
| 374 |
|
Pcleft = kt0* y* M ./ (y*(l+r)+ kt0* l);
|
| 375 |
|
Pavailable = Pcleft*(l+r)./kt0;
|
| 376 |
|
Preprocess = Pcleft*r/x; % canbe fractional
|
| 377 |
|
|
| 378 |
|
ANprobability=zeros(nANchannels,segmentLength);
|
| 379 |
|
ANprobRateOutput=zeros(nANchannels,signalLength);
|
| 380 |
|
lengthAbsRefractoryP= round(AN_refractory_period/dt);
|
| 381 |
|
% special variables for monitoring synaptic cleft (specialists only)
|
| 382 |
|
savePavailableSeg=zeros(nANchannels,segmentLength);
|
| 383 |
|
savePavailable=zeros(nANchannels,signalLength);
|
| 384 |
|
|
| 385 |
|
% spikes % ! ! ! ! ! ! ! !
|
| 386 |
|
lengthAbsRefractory= round(AN_refractory_period/ANdt);
|
| 387 |
|
|
| 388 |
|
AN_ydt= repmat(AN_IHCsynapseParams.y*ANdt, nANfibers,1);
|
| 389 |
|
AN_ldt= repmat(AN_IHCsynapseParams.l*ANdt, nANfibers,1);
|
| 390 |
|
AN_xdt= repmat(AN_IHCsynapseParams.x*ANdt, nANfibers,1);
|
| 391 |
|
AN_rdt= repmat(AN_IHCsynapseParams.r*ANdt, nANfibers,1);
|
| 392 |
|
AN_rdt_plus_ldt= AN_rdt + AN_ldt;
|
| 393 |
|
AN_M= round(AN_IHCsynapseParams.M);
|
| 394 |
|
|
| 395 |
|
% kt0 is initial release rate
|
| 396 |
|
% Establish as a vector (length=channel x number of fibers)
|
| 397 |
|
kt0= repmat(kt0', nFibersPerChannel, 1);
|
| 398 |
|
kt0=reshape(kt0, nANfibers,1);
|
| 399 |
|
|
| 400 |
|
% starting values for reservoirs
|
| 401 |
|
AN_cleft = kt0* y* M ./ (y*(l+r)+ kt0* l);
|
| 402 |
|
AN_available = round(AN_cleft*(l+r)./kt0); %must be integer
|
| 403 |
|
AN_reprocess = AN_cleft*r/x;
|
| 404 |
|
|
| 405 |
|
% output is in a logical array spikes = 1/ 0.
|
| 406 |
|
ANspikes= false(nANfibers,reducedSegmentLength);
|
| 407 |
|
ANoutput= false(nANfibers,reducedSignalLength);
|
| 408 |
|
|
| 409 |
|
|
| 410 |
|
%% CN (first brain stem nucleus - could be any subdivision of CN)
|
| 411 |
|
% Input to a CN neuorn is a random selection of AN fibers within a channel
|
| 412 |
|
% The number of AN fibers used is ANfibersFanInToCN
|
| 413 |
|
% CNtauGk (Potassium time constant) determines the rate of firing of
|
| 414 |
|
% the unit when driven hard by a DC input (not normally >350 sp/s)
|
| 415 |
|
% If there is more than one value, everything is replicated accordingly
|
| 416 |
|
|
| 417 |
|
ANavailableFibersPerChan=AN_IHCsynapseParams.numFibers;
|
| 418 |
|
ANfibersFanInToCN=MacGregorMultiParams.fibersPerNeuron;
|
| 419 |
|
|
| 420 |
|
CNtauGk=MacGregorMultiParams.tauGk; % row vector of CN types (by tauGk)
|
| 421 |
|
nCNtauGk=length(CNtauGk);
|
| 422 |
|
|
| 423 |
|
% the total number of 'channels' is now greater
|
| 424 |
|
nCNchannels=nANchannels*nCNtauGk;
|
| 425 |
|
|
| 426 |
|
nCNneuronsPerChannel=MacGregorMultiParams.nNeuronsPerBF;
|
| 427 |
|
tauGk=repmat(CNtauGk, nCNneuronsPerChannel,1);
|
| 428 |
|
tauGk=reshape(tauGk,nCNneuronsPerChannel*nCNtauGk,1);
|
| 429 |
|
|
| 430 |
|
% Now the number of neurons has been increased
|
| 431 |
|
nCNneurons=nCNneuronsPerChannel*nCNchannels;
|
| 432 |
|
CNmembranePotential=zeros(nCNneurons,reducedSegmentLength);
|
| 433 |
|
|
| 434 |
|
% establish which ANfibers (by name) feed into which CN nuerons
|
| 435 |
|
CNinputfiberLists=zeros(nANchannels*nCNneuronsPerChannel, ANfibersFanInToCN);
|
| 436 |
|
unitNo=1;
|
| 437 |
|
for ch=1:nANchannels
|
| 438 |
|
% Each channel contains a number of units =length(listOfFanInValues)
|
| 439 |
|
for idx=1:nCNneuronsPerChannel
|
| 440 |
|
for idx2=1:nCNtauGk
|
| 441 |
|
fibersUsed=(ch-1)*ANavailableFibersPerChan + ...
|
| 442 |
|
ceil(rand(1,ANfibersFanInToCN)* ANavailableFibersPerChan);
|
| 443 |
|
CNinputfiberLists(unitNo,:)=fibersUsed;
|
| 444 |
|
unitNo=unitNo+1;
|
| 445 |
|
end
|
| 446 |
|
end
|
| 447 |
|
end
|
| 448 |
|
|
| 449 |
|
% input to CN units
|
| 450 |
|
AN_PSTH=zeros(nCNneurons,reducedSegmentLength);
|
| 451 |
|
|
| 452 |
|
% Generate CNalphaFunction function
|
| 453 |
|
% by which spikes are converted to post-synaptic currents
|
| 454 |
|
CNdendriteLPfreq= MacGregorMultiParams.dendriteLPfreq;
|
| 455 |
|
CNcurrentPerSpike=MacGregorMultiParams.currentPerSpike;
|
| 456 |
|
CNspikeToCurrentTau=1/(2*pi*CNdendriteLPfreq);
|
| 457 |
|
t=ANdt:ANdt:5*CNspikeToCurrentTau;
|
| 458 |
|
CNalphaFunction= (1 / ...
|
| 459 |
|
CNspikeToCurrentTau)*t.*exp(-t /CNspikeToCurrentTau);
|
| 460 |
|
CNalphaFunction=CNalphaFunction*CNcurrentPerSpike;
|
| 461 |
|
|
| 462 |
|
% figure(98), plot(t,CNalphaFunction)
|
| 463 |
|
% working memory for implementing convolution
|
| 464 |
|
|
| 465 |
|
CNcurrentTemp=...
|
| 466 |
|
zeros(nCNneurons,reducedSegmentLength+length(CNalphaFunction)-1);
|
| 467 |
|
% trailing alphas are parts of humps carried forward to the next segment
|
| 468 |
|
CNtrailingAlphas=zeros(nCNneurons,length(CNalphaFunction));
|
| 469 |
|
|
| 470 |
|
CN_tauM=MacGregorMultiParams.tauM;
|
| 471 |
|
CN_tauTh=MacGregorMultiParams.tauTh;
|
| 472 |
|
CN_cap=MacGregorMultiParams.Cap;
|
| 473 |
|
CN_c=MacGregorMultiParams.c;
|
| 474 |
|
CN_b=MacGregorMultiParams.dGkSpike;
|
| 475 |
|
CN_Ek=MacGregorMultiParams.Ek;
|
| 476 |
|
CN_Eb= MacGregorMultiParams.Eb;
|
| 477 |
|
CN_Er=MacGregorMultiParams.Er;
|
| 478 |
|
CN_Th0= MacGregorMultiParams.Th0;
|
| 479 |
|
CN_E= zeros(nCNneurons,1);
|
| 480 |
|
CN_Gk= zeros(nCNneurons,1);
|
| 481 |
|
CN_Th= MacGregorMultiParams.Th0*ones(nCNneurons,1);
|
| 482 |
|
CN_Eb=CN_Eb.*ones(nCNneurons,1);
|
| 483 |
|
CN_Er=CN_Er.*ones(nCNneurons,1);
|
| 484 |
|
CNtimeSinceLastSpike=zeros(nCNneurons,1);
|
| 485 |
|
% tauGk is the main distinction between neurons
|
| 486 |
|
% in fact they are all the same in the standard model
|
| 487 |
|
tauGk=repmat(tauGk,nANchannels,1);
|
| 488 |
|
|
| 489 |
|
CNoutput=false(nCNneurons,reducedSignalLength);
|
| 490 |
|
|
| 491 |
|
|
| 492 |
|
%% MacGregor (IC - second nucleus) --------
|
| 493 |
|
nICcells=nANchannels*nCNtauGk; % one cell per channel
|
| 494 |
|
CN_PSTH=zeros(nICcells ,reducedSegmentLength);
|
| 495 |
|
|
| 496 |
|
ICspikeWidth=0.00015; % this may need revisiting
|
| 497 |
|
epochsPerSpike=round(ICspikeWidth/ANdt);
|
| 498 |
|
if epochsPerSpike<1
|
| 499 |
|
error(['MacGregorMulti: sample rate too low to support ' ...
|
| 500 |
|
num2str(ICspikeWidth*1e6) ' microsec spikes']);
|
| 501 |
|
end
|
| 502 |
|
|
| 503 |
|
% short names
|
| 504 |
|
IC_tauM=MacGregorParams.tauM;
|
| 505 |
|
IC_tauGk=MacGregorParams.tauGk;
|
| 506 |
|
IC_tauTh=MacGregorParams.tauTh;
|
| 507 |
|
IC_cap=MacGregorParams.Cap;
|
| 508 |
|
IC_c=MacGregorParams.c;
|
| 509 |
|
IC_b=MacGregorParams.dGkSpike;
|
| 510 |
|
IC_Th0=MacGregorParams.Th0;
|
| 511 |
|
IC_Ek=MacGregorParams.Ek;
|
| 512 |
|
IC_Eb= MacGregorParams.Eb;
|
| 513 |
|
IC_Er=MacGregorParams.Er;
|
| 514 |
|
|
| 515 |
|
IC_E=zeros(nICcells,1);
|
| 516 |
|
IC_Gk=zeros(nICcells,1);
|
| 517 |
|
IC_Th=IC_Th0*ones(nICcells,1);
|
| 518 |
|
|
| 519 |
|
% Dendritic filtering, all spikes are replaced by CNalphaFunction functions
|
| 520 |
|
ICdendriteLPfreq= MacGregorParams.dendriteLPfreq;
|
| 521 |
|
ICcurrentPerSpike=MacGregorParams.currentPerSpike;
|
| 522 |
|
ICspikeToCurrentTau=1/(2*pi*ICdendriteLPfreq);
|
| 523 |
|
t=ANdt:ANdt:3*ICspikeToCurrentTau;
|
| 524 |
|
IC_CNalphaFunction= (ICcurrentPerSpike / ...
|
| 525 |
|
ICspikeToCurrentTau)*t.*exp(-t / ICspikeToCurrentTau);
|
| 526 |
|
% figure(98), plot(t,IC_CNalphaFunction)
|
| 527 |
|
|
| 528 |
|
% working space for implementing alpha function
|
| 529 |
|
ICcurrentTemp=...
|
| 530 |
|
zeros(nICcells,reducedSegmentLength+length(IC_CNalphaFunction)-1);
|
| 531 |
|
ICtrailingAlphas=zeros(nICcells, length(IC_CNalphaFunction));
|
| 532 |
|
|
| 533 |
|
ICfiberTypeRates=zeros(nANfiberTypes,reducedSignalLength);
|
| 534 |
|
ICoutput=false(nICcells,reducedSignalLength);
|
| 535 |
|
|
| 536 |
|
ICmembranePotential=zeros(nICcells,reducedSegmentLength);
|
| 537 |
|
ICmembraneOutput=zeros(nICcells,signalLength);
|
| 538 |
|
|
| 539 |
|
|
| 540 |
|
%% Main program %% %% %% %% %% %% %% %% %% %% %% %% %% %%
|
| 541 |
|
|
| 542 |
|
% Compute the entire model for each segment
|
| 543 |
|
segmentStartPTR=1;
|
| 544 |
|
reducedSegmentPTR=1; % when sampling rate is reduced
|
| 545 |
|
while segmentStartPTR<signalLength
|
| 546 |
|
segmentEndPTR=segmentStartPTR+segmentLength-1;
|
| 547 |
|
% shorter segments after speed up.
|
| 548 |
|
shorterSegmentEndPTR=reducedSegmentPTR+reducedSegmentLength-1;
|
| 549 |
|
|
| 550 |
|
inputPressureSegment=inputSignal...
|
| 551 |
|
(:,segmentStartPTR:segmentStartPTR+segmentLength-1);
|
| 552 |
|
|
| 553 |
|
% segment debugging plots
|
| 554 |
|
% figure(98)
|
| 555 |
|
% plot(segmentTime,inputPressureSegment), title('signalSegment')
|
| 556 |
|
|
| 557 |
|
|
| 558 |
|
% OME ----------------------
|
| 559 |
|
|
| 560 |
|
% OME Stage 1: external resonances. Add to inputSignal pressure wave
|
| 561 |
|
y=inputPressureSegment;
|
| 562 |
|
for n=1:nOMEExtFilters
|
| 563 |
|
% any number of resonances can be used
|
| 564 |
|
[x OMEExtFilterBndry{n}] = ...
|
| 565 |
|
filter(ExtFilter_b{n},ExtFilter_a{n},...
|
| 566 |
|
inputPressureSegment, OMEExtFilterBndry{n});
|
| 567 |
|
x= x* OMEgainScalars(n);
|
| 568 |
|
% This is a parallel resonance so add it
|
| 569 |
|
y=y+x;
|
| 570 |
|
end
|
| 571 |
|
inputPressureSegment=y;
|
| 572 |
|
OMEextEarPressure(segmentStartPTR:segmentEndPTR)= inputPressureSegment;
|
| 573 |
|
|
| 574 |
|
% OME stage 2: convert input pressure (velocity) to
|
| 575 |
|
% tympanic membrane(TM) displacement using low pass filter
|
| 576 |
|
[TMdisplacementSegment OME_TMdisplacementBndry] = ...
|
| 577 |
|
filter(TMdisp_b,TMdisp_a,inputPressureSegment, ...
|
| 578 |
|
OME_TMdisplacementBndry);
|
| 579 |
|
% and save it
|
| 580 |
|
TMoutput(segmentStartPTR:segmentEndPTR)= TMdisplacementSegment;
|
| 581 |
|
|
| 582 |
|
% OME stage 3: middle ear high pass effect to simulate stapes inertia
|
| 583 |
|
[stapesDisplacement OMEhighPassBndry] = ...
|
| 584 |
|
filter(stapesDisp_b,stapesDisp_a,TMdisplacementSegment, ...
|
| 585 |
|
OMEhighPassBndry);
|
| 586 |
|
|
| 587 |
|
% OME stage 4: apply stapes scalar
|
| 588 |
|
stapesDisplacement=stapesDisplacement*stapesScalar;
|
| 589 |
|
|
| 590 |
|
% OME stage 5: acoustic reflex stapes attenuation
|
| 591 |
|
% Attenuate the TM response using feedback from LSR fiber activity
|
| 592 |
|
if segmentStartPTR>efferentDelayPts
|
| 593 |
|
stapesDisplacement= stapesDisplacement.*...
|
| 594 |
|
ARattenuation(segmentStartPTR-efferentDelayPts:...
|
| 595 |
|
segmentEndPTR-efferentDelayPts);
|
| 596 |
|
end
|
| 597 |
|
|
| 598 |
|
% segment debugging plots
|
| 599 |
|
% figure(98)
|
| 600 |
|
% plot(segmentTime, stapesDisplacement), title ('stapesDisplacement')
|
| 601 |
|
|
| 602 |
|
% and save
|
| 603 |
|
OMEoutput(segmentStartPTR:segmentEndPTR)= stapesDisplacement;
|
| 604 |
|
|
| 605 |
|
|
| 606 |
|
%% BM ------------------------------
|
| 607 |
|
% Each location is computed separately
|
| 608 |
|
for BFno=1:nBFs
|
| 609 |
|
|
| 610 |
|
% *linear* path
|
| 611 |
|
linOutput = stapesDisplacement * linGAIN; % linear gain
|
| 612 |
|
for order = 1 : GTlinOrder
|
| 613 |
|
[linOutput GTlinBdry{BFno,order}] = ...
|
| 614 |
|
filter(GTlin_b(BFno,:), GTlin_a(BFno,:), linOutput, GTlinBdry{BFno,order});
|
| 615 |
|
end
|
| 616 |
|
|
| 617 |
|
% *nonLinear* path
|
| 618 |
|
% efferent attenuation (0 <> 1)
|
| 619 |
|
if segmentStartPTR>efferentDelayPts
|
| 620 |
|
MOC=MOCattenuation(BFno, segmentStartPTR-efferentDelayPts:...
|
| 621 |
|
segmentEndPTR-efferentDelayPts);
|
| 622 |
|
else % no MOC available yet
|
| 623 |
|
MOC=ones(1, segmentLength);
|
| 624 |
|
end
|
| 625 |
|
% apply MOC to nonlinear input function
|
| 626 |
|
nonlinOutput=stapesDisplacement.* MOC;
|
| 627 |
|
|
| 628 |
|
% first gammatone filter (nonlin path)
|
| 629 |
|
for order = 1 : GTnonlinOrder
|
| 630 |
|
[nonlinOutput GTnonlinBdry1{BFno,order}] = ...
|
| 631 |
|
filter(GTnonlin_b(BFno,:), GTnonlin_a(BFno,:), ...
|
| 632 |
|
nonlinOutput, GTnonlinBdry1{BFno,order});
|
| 633 |
|
end
|
| 634 |
|
% broken stick instantaneous compression
|
| 635 |
|
y= nonlinOutput.* DRNLa; % linear section.
|
| 636 |
|
% compress parts of the signal above the compression threshold
|
| 637 |
|
abs_x = abs(nonlinOutput);
|
| 638 |
|
idx=find(abs_x>DRNLcompressionThreshold);
|
| 639 |
|
if ~isempty(idx)>0
|
| 640 |
|
y(idx)=sign(y(idx)).* (DRNLb*abs_x(idx).^DRNLc);
|
| 641 |
|
end
|
| 642 |
|
nonlinOutput=y;
|
| 643 |
|
|
| 644 |
|
% second filter removes distortion products
|
| 645 |
|
for order = 1 : GTnonlinOrder
|
| 646 |
|
[ nonlinOutput GTnonlinBdry2{BFno,order}] = ...
|
| 647 |
|
filter(GTnonlin_b(BFno,:), GTnonlin_a(BFno,:), ...
|
| 648 |
|
nonlinOutput, GTnonlinBdry2{BFno,order});
|
| 649 |
|
end
|
| 650 |
|
|
| 651 |
|
% combine the two paths to give the DRNL displacement
|
| 652 |
|
DRNLresponse(BFno,:)=linOutput+nonlinOutput;
|
| 653 |
|
end % BF
|
| 654 |
|
|
| 655 |
|
% segment debugging plots
|
| 656 |
|
% figure(98)
|
| 657 |
|
% if size(DRNLresponse,1)>3
|
| 658 |
|
% imagesc(DRNLresponse) % matrix display
|
| 659 |
|
% title('DRNLresponse'); % single or double channel response
|
| 660 |
|
% else
|
| 661 |
|
% plot(segmentTime, DRNLresponse)
|
| 662 |
|
% end
|
| 663 |
|
|
| 664 |
|
% and save it
|
| 665 |
|
DRNLoutput(:, segmentStartPTR:segmentEndPTR)= DRNLresponse;
|
| 666 |
|
|
| 667 |
|
|
| 668 |
|
%% IHC ------------------------------------
|
| 669 |
|
% BM displacement to IHCciliaDisplacement is a high-pass filter
|
| 670 |
|
% because of viscous coupling
|
| 671 |
|
for idx=1:nBFs
|
| 672 |
|
[IHCciliaDisplacement(idx,:) IHCciliaBndry{idx}] = ...
|
| 673 |
|
filter(IHCciliaFilter_b,IHCciliaFilter_a, ...
|
| 674 |
|
DRNLresponse(idx,:), IHCciliaBndry{idx});
|
| 675 |
|
end
|
| 676 |
|
|
| 677 |
|
% apply scalar
|
| 678 |
|
IHCciliaDisplacement=IHCciliaDisplacement* IHC_C;
|
| 679 |
|
|
| 680 |
|
% and save it
|
| 681 |
|
IHC_cilia_output(:,segmentStartPTR:segmentStartPTR+segmentLength-1)=...
|
| 682 |
|
IHCciliaDisplacement;
|
| 683 |
|
|
| 684 |
|
% compute apical conductance
|
| 685 |
|
G=IHCGmax./(1+exp(-(IHCciliaDisplacement-IHCu0)/IHCs0).*...
|
| 686 |
|
(1+exp(-(IHCciliaDisplacement-IHCu1)/IHCs1)));
|
| 687 |
|
Gu=G + IHCGa;
|
| 688 |
|
|
| 689 |
|
% Compute receptor potential
|
| 690 |
|
for idx=1:segmentLength
|
| 691 |
|
IHC_Vnow=IHC_Vnow+ (-Gu(:, idx).*(IHC_Vnow-IHC_Et)-...
|
| 692 |
|
IHC_Gk*(IHC_Vnow-IHC_Ekp))* dt/IHC_Cab;
|
| 693 |
|
IHC_RP(:,idx)=IHC_Vnow;
|
| 694 |
|
end
|
| 695 |
|
|
| 696 |
|
% segment debugging plots
|
| 697 |
|
% if size(IHC_RP,1)>3
|
| 698 |
|
% surf(IHC_RP), shading interp, title('IHC_RP')
|
| 699 |
|
% else
|
| 700 |
|
% plot(segmentTime, IHC_RP)
|
| 701 |
|
% end
|
| 702 |
|
|
| 703 |
|
% and save it
|
| 704 |
|
IHCoutput(:, segmentStartPTR:segmentStartPTR+segmentLength-1)=IHC_RP;
|
| 705 |
|
|
| 706 |
|
|
| 707 |
|
%% synapse -----------------------------
|
| 708 |
|
% Compute the vesicle release rate for each fiber type at each BF
|
| 709 |
|
% replicate IHC_RP for each fiber type
|
| 710 |
|
Vsynapse=repmat(IHC_RP, nANfiberTypes,1);
|
| 711 |
|
|
| 712 |
|
% look-up table of target fraction channels open for a given IHC_RP
|
| 713 |
|
mICaINF= 1./( 1 + exp(-gamma * Vsynapse) /beta);
|
| 714 |
|
% fraction of channel open - apply time constant
|
| 715 |
|
for idx=1:segmentLength
|
| 716 |
|
% mICaINF is the current 'target' value of mICa
|
| 717 |
|
mICaCurrent=mICaCurrent+(mICaINF(:,idx)-mICaCurrent)*dt./tauM;
|
| 718 |
|
mICa(:,idx)=mICaCurrent;
|
| 719 |
|
end
|
| 720 |
|
|
| 721 |
|
ICa= (GmaxCa* mICa.^3) .* (Vsynapse- ECa);
|
| 722 |
|
|
| 723 |
|
for idx=1:segmentLength
|
| 724 |
|
CaCurrent=CaCurrent + ICa(:,idx)*dt - CaCurrent*dt./tauCa;
|
| 725 |
|
synapticCa(:,idx)=CaCurrent;
|
| 726 |
|
end
|
| 727 |
|
synapticCa=-synapticCa; % treat IHCpreSynapseParams as positive substance
|
| 728 |
|
|
| 729 |
|
% NB vesicleReleaseRate is /s and is independent of dt
|
| 730 |
|
vesicleReleaseRate = synapse_z * synapticCa.^synapse_power; % rate
|
| 731 |
|
|
| 732 |
|
% segment debugging plots
|
| 733 |
|
% if size(vesicleReleaseRate,1)>3
|
| 734 |
|
% surf(vesicleReleaseRate), shading interp, title('vesicleReleaseRate')
|
| 735 |
|
% else
|
| 736 |
|
% plot(segmentTime, vesicleReleaseRate)
|
| 737 |
|
% end
|
| 738 |
|
|
| 739 |
|
|
| 740 |
|
%% AN
|
| 741 |
|
switch AN_spikesOrProbability
|
| 742 |
|
case 'probability'
|
| 743 |
|
% No refractory effect is applied
|
| 744 |
|
for t = 1:segmentLength;
|
| 745 |
|
M_Pq=PAN_M-Pavailable;
|
| 746 |
|
M_Pq(M_Pq<0)=0;
|
| 747 |
|
Preplenish = M_Pq .* PAN_ydt;
|
| 748 |
|
Pejected = Pavailable.* vesicleReleaseRate(:,t)*dt;
|
| 749 |
|
Preprocessed = M_Pq.*Preprocess.* PAN_xdt;
|
| 750 |
|
|
| 751 |
|
ANprobability(:,t)= min(Pejected,1);
|
| 752 |
|
reuptakeandlost= PAN_rdt_plus_ldt .* Pcleft;
|
| 753 |
|
reuptake= PAN_rdt.* Pcleft;
|
| 754 |
|
|
| 755 |
|
Pavailable= Pavailable+ Preplenish- Pejected+ Preprocessed;
|
| 756 |
|
Pcleft= Pcleft + Pejected - reuptakeandlost;
|
| 757 |
|
Preprocess= Preprocess + reuptake - Preprocessed;
|
| 758 |
|
Pavailable(Pavailable<0)=0;
|
| 759 |
|
savePavailableSeg(:,t)=Pavailable; % synapse tracking
|
| 760 |
|
end
|
| 761 |
|
% and save it as *rate*
|
| 762 |
|
ANrate=ANprobability/dt;
|
| 763 |
|
ANprobRateOutput(:, segmentStartPTR:...
|
| 764 |
|
segmentStartPTR+segmentLength-1)= ANrate;
|
| 765 |
|
% monitor synapse contents (only sometimes used)
|
| 766 |
|
savePavailable(:, segmentStartPTR:segmentStartPTR+segmentLength-1)=...
|
| 767 |
|
savePavailableSeg;
|
| 768 |
|
|
| 769 |
|
% Estimate efferent effects. ARattenuation (0 <> 1)
|
| 770 |
|
% acoustic reflex
|
| 771 |
|
[r c]=size(ANrate);
|
| 772 |
|
if r>nBFs % Only if LSR fibers are computed
|
| 773 |
|
ARAttSeg=mean(ANrate(1:nBFs,:),1); %LSR channels are 1:nBF
|
| 774 |
|
% smooth
|
| 775 |
|
[ARAttSeg, ARboundaryProb] = ...
|
| 776 |
|
filter(ARfilt_b, ARfilt_a, ARAttSeg, ARboundaryProb);
|
| 777 |
|
ARAttSeg=ARAttSeg-ARrateThreshold;
|
| 778 |
|
ARAttSeg(ARAttSeg<0)=0; % prevent negative strengths
|
| 779 |
|
ARattenuation(segmentStartPTR:segmentEndPTR)=...
|
| 780 |
|
(1-ARrateToAttenuationFactorProb.* ARAttSeg);
|
| 781 |
|
end
|
| 782 |
|
% plot(ARattenuation)
|
| 783 |
|
|
| 784 |
|
% MOC attenuation
|
| 785 |
|
% within-channel HSR response only
|
| 786 |
|
HSRbegins=nBFs*(nANfiberTypes-1)+1;
|
| 787 |
|
rates=ANrate(HSRbegins:end,:);
|
| 788 |
|
if rateToAttenuationFactorProb<0
|
| 789 |
|
% negative factor implies a fixed attenuation
|
| 790 |
|
MOCattenuation(:,segmentStartPTR:segmentEndPTR)= ...
|
| 791 |
|
ones(size(rates))* -rateToAttenuationFactorProb;
|
| 792 |
|
else
|
| 793 |
|
for idx=1:nBFs
|
| 794 |
|
[smoothedRates, MOCprobBoundary{idx}] = ...
|
| 795 |
|
filter(MOCfilt_b, MOCfilt_a, rates(idx,:), ...
|
| 796 |
|
MOCprobBoundary{idx});
|
| 797 |
|
smoothedRates=smoothedRates-MOCrateThresholdProb;
|
| 798 |
|
smoothedRates(smoothedRates<0)=0;
|
| 799 |
|
MOCattenuation(idx,segmentStartPTR:segmentEndPTR)= ...
|
| 800 |
|
(1- smoothedRates* rateToAttenuationFactorProb);
|
| 801 |
|
end
|
| 802 |
|
end
|
| 803 |
|
MOCattenuation(MOCattenuation<0)=0.001;
|
| 804 |
|
|
| 805 |
|
% plot(MOCattenuation)
|
| 806 |
|
|
| 807 |
|
|
| 808 |
|
case 'spikes'
|
| 809 |
|
ANtimeCount=0;
|
| 810 |
|
% implement speed upt
|
| 811 |
|
for t = ANspeedUpFactor:ANspeedUpFactor:segmentLength;
|
| 812 |
|
ANtimeCount=ANtimeCount+1;
|
| 813 |
|
% convert release rate to probabilities
|
| 814 |
|
releaseProb=vesicleReleaseRate(:,t)*ANdt;
|
| 815 |
|
% releaseProb is the release probability per channel
|
| 816 |
|
% but each channel has many synapses
|
| 817 |
|
releaseProb=repmat(releaseProb',nFibersPerChannel,1);
|
| 818 |
|
releaseProb=reshape(releaseProb, nFibersPerChannel*nANchannels,1);
|
| 819 |
|
|
| 820 |
|
% AN_available=round(AN_available); % vesicles must be integer, (?needed)
|
| 821 |
|
M_q=AN_M- AN_available; % number of missing vesicles
|
| 822 |
|
M_q(M_q<0)= 0; % cannot be less than 0
|
| 823 |
|
|
| 824 |
|
% AN_N1 converts probability to discrete events
|
| 825 |
|
% it considers each event that might occur
|
| 826 |
|
% (how many vesicles might be released)
|
| 827 |
|
% and returns a count of how many were released
|
| 828 |
|
|
| 829 |
|
% slow line
|
| 830 |
|
% probabilities= 1-(1-releaseProb).^AN_available;
|
| 831 |
|
probabilities= 1-intpow((1-releaseProb), AN_available);
|
| 832 |
|
ejected= probabilities> rand(length(AN_available),1);
|
| 833 |
|
|
| 834 |
|
reuptakeandlost = AN_rdt_plus_ldt .* AN_cleft;
|
| 835 |
|
reuptake = AN_rdt.* AN_cleft;
|
| 836 |
|
|
| 837 |
|
% slow line
|
| 838 |
|
% probabilities= 1-(1-AN_reprocess.*AN_xdt).^M_q;
|
| 839 |
|
probabilities= 1-intpow((1-AN_reprocess.*AN_xdt), M_q);
|
| 840 |
|
reprocessed= probabilities>rand(length(M_q),1);
|
| 841 |
|
|
| 842 |
|
% slow line
|
| 843 |
|
% probabilities= 1-(1-AN_ydt).^M_q;
|
| 844 |
|
probabilities= 1-intpow((1-AN_ydt), M_q);
|
| 845 |
|
|
| 846 |
|
replenish= probabilities>rand(length(M_q),1);
|
| 847 |
|
|
| 848 |
|
AN_available = AN_available + replenish - ejected ...
|
| 849 |
|
+ reprocessed;
|
| 850 |
|
AN_cleft = AN_cleft + ejected - reuptakeandlost;
|
| 851 |
|
AN_reprocess = AN_reprocess + reuptake - reprocessed;
|
| 852 |
|
|
| 853 |
|
% ANspikes is logical record of vesicle release events>0
|
| 854 |
|
ANspikes(:, ANtimeCount)= ejected;
|
| 855 |
|
end % t
|
| 856 |
|
|
| 857 |
|
% zero any events that are preceded by release events ...
|
| 858 |
|
% within the refractory period
|
| 859 |
|
% The refractory period consist of two periods
|
| 860 |
|
% 1) the absolute period where no spikes occur
|
| 861 |
|
% 2) a relative period where a spike may occur. This relative
|
| 862 |
|
% period is realised as a variable length interval
|
| 863 |
|
% where the length is chosen at random
|
| 864 |
|
% (esentially a linear ramp up)
|
| 865 |
|
|
| 866 |
|
% Andreas has a fix for this
|
| 867 |
|
for t = 1:ANtimeCount-2*lengthAbsRefractory;
|
| 868 |
|
% identify all spikes across fiber array at time (t)
|
| 869 |
|
% idx is a list of channels where spikes occurred
|
| 870 |
|
% ?? try sparse matrices?
|
| 871 |
|
idx=find(ANspikes(:,t));
|
| 872 |
|
for j=idx % consider each spike
|
| 873 |
|
% specify variable refractory period
|
| 874 |
|
% between abs and 2*abs refractory period
|
| 875 |
|
nPointsRefractory=lengthAbsRefractory+...
|
| 876 |
|
round(rand*lengthAbsRefractory);
|
| 877 |
|
% disable spike potential for refractory period
|
| 878 |
|
% set all values in this range to 0
|
| 879 |
|
ANspikes(j,t+1:t+nPointsRefractory)=0;
|
| 880 |
|
end
|
| 881 |
|
end %t
|
| 882 |
|
|
| 883 |
|
% segment debugging
|
| 884 |
|
% plotInstructions.figureNo=98;
|
| 885 |
|
% plotInstructions.displaydt=ANdt;
|
| 886 |
|
% plotInstructions.numPlots=1;
|
| 887 |
|
% plotInstructions.subPlotNo=1;
|
| 888 |
|
% UTIL_plotMatrix(ANspikes, plotInstructions);
|
| 889 |
|
|
| 890 |
|
% and save it. NB, AN is now on 'speedUp' time
|
| 891 |
|
ANoutput(:, reducedSegmentPTR: shorterSegmentEndPTR)=ANspikes;
|
| 892 |
|
|
| 893 |
|
|
| 894 |
|
%% CN Macgregor first neucleus -------------------------------
|
| 895 |
|
% input is from AN so ANdt is used throughout
|
| 896 |
|
% Each CNneuron has a unique set of input fibers selected
|
| 897 |
|
% at random from the available AN fibers (CNinputfiberLists)
|
| 898 |
|
|
| 899 |
|
% Create the dendritic current for that neuron
|
| 900 |
|
% First get input spikes to this neuron
|
| 901 |
|
synapseNo=1;
|
| 902 |
|
for ch=1:nCNchannels
|
| 903 |
|
for idx=1:nCNneuronsPerChannel
|
| 904 |
|
% determine candidate fibers for this unit
|
| 905 |
|
fibersUsed=CNinputfiberLists(synapseNo,:);
|
| 906 |
|
% ANpsth has a bin width of ANdt
|
| 907 |
|
% (just a simple sum across fibers)
|
| 908 |
|
AN_PSTH(synapseNo,:) = ...
|
| 909 |
|
sum(ANspikes(fibersUsed,:), 1);
|
| 910 |
|
synapseNo=synapseNo+1;
|
| 911 |
|
end
|
| 912 |
|
end
|
| 913 |
|
|
| 914 |
|
|
| 915 |
|
|
| 916 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
| 917 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
| 918 |
|
|
| 919 |
|
|
| 920 |
|
|
| 921 |
|
% One alpha function per spike
|
| 922 |
|
[alphaRows alphaCols]=size(CNtrailingAlphas);
|
| 923 |
|
|
| 924 |
|
for unitNo=1:nCNneurons
|
| 925 |
|
|
| 926 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
| 927 |
|
|
| 928 |
|
CNcurrentTemp0(unitNo,:)= ...
|
| 929 |
|
conv(AN_PSTH(unitNo,:),CNalphaFunction);
|
| 930 |
|
|
| 931 |
|
|
| 932 |
|
|
| 933 |
|
CNcurrentTemp(unitNo,:)= ...
|
| 934 |
|
conv2(AN_PSTH(unitNo,:),CNalphaFunction);
|
| 935 |
|
% Changed conv to conv2 because it runs faster. (Andreas)
|
| 936 |
|
|
| 937 |
|
|
| 938 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
| 939 |
|
|
| 940 |
|
%
|
| 941 |
|
%
|
| 942 |
|
% f = CNalphaFunction;
|
| 943 |
|
% g = AN_PSTH(unitNo,:);
|
| 944 |
|
%
|
| 945 |
|
%
|
| 946 |
|
% g = [g zeros(1,length(f)-1)];
|
| 947 |
|
%
|
| 948 |
|
% spikePos = find(g)';
|
| 949 |
|
%
|
| 950 |
|
% result = zeros(1,length(g));
|
| 951 |
|
%
|
| 952 |
|
% for index = 1:length(spikePos)
|
| 953 |
|
% k = spikePos(index);
|
| 954 |
|
% result(k:(k+length(f)-1)) = result(k:(k+length(f)-1)) + g(k)*f;
|
| 955 |
|
% end
|
| 956 |
|
%
|
| 957 |
|
% CNcurrentTemp2(unitNo,:) = result;
|
| 958 |
|
|
| 959 |
|
|
| 960 |
|
end
|
| 961 |
|
|
| 962 |
|
|
| 963 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
| 964 |
|
|
| 965 |
|
|
| 966 |
|
|
| 967 |
|
|
| 968 |
|
f = CNalphaFunction;
|
| 969 |
|
g = AN_PSTH;
|
| 970 |
|
|
| 971 |
|
g = [g zeros(size(g,1),length(f)-1)];
|
| 972 |
|
|
| 973 |
|
[r c] = find(g);
|
| 974 |
|
|
| 975 |
|
CNcurrentTemp2 = zeros(size(g));
|
| 976 |
|
|
| 977 |
|
for index = 1:length(r)
|
| 978 |
|
|
| 979 |
|
row = r(index);
|
| 980 |
|
col = c(index);
|
| 981 |
|
|
| 982 |
|
CNcurrentTemp2(row,col:col+length(f)-1) = CNcurrentTemp2(row,col:col+length(f)-1) + f*g(row,col);
|
| 983 |
|
|
| 984 |
|
end
|
| 985 |
|
|
| 986 |
|
|
| 987 |
|
|
| 988 |
|
CONVOLUTION_CHANGE_TEST = CONVOLUTION_CHANGE_TEST + sum(abs(CNcurrentTemp2 - CNcurrentTemp))+ sum(abs(CNcurrentTemp0 - CNcurrentTemp));
|
| 989 |
|
|
| 990 |
|
|
| 991 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
| 992 |
|
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
| 993 |
|
|
| 994 |
|
|
| 995 |
|
|
| 996 |
|
|
| 997 |
|
% disp(['sum(AN_PSTH)= ' num2str(sum(AN_PSTH(1,:)))])
|
| 998 |
|
% add post-synaptic current left over from previous segment
|
| 999 |
|
CNcurrentTemp(:,1:alphaCols)=...
|
| 1000 |
|
CNcurrentTemp(:,1:alphaCols)+ CNtrailingAlphas;
|
| 1001 |
|
|
| 1002 |
|
% take post-synaptic current for this segment
|
| 1003 |
|
CNcurrentInput= CNcurrentTemp(:, 1:reducedSegmentLength);
|
| 1004 |
|
% disp(['mean(CNcurrentInput)= ' num2str(mean(CNcurrentInput(1,:)))])
|
| 1005 |
|
|
| 1006 |
|
% trailingalphas are the ends of the alpha functions that
|
| 1007 |
|
% spill over into the next segment
|
| 1008 |
|
CNtrailingAlphas= ...
|
| 1009 |
|
CNcurrentTemp(:, reducedSegmentLength+1:end);
|
| 1010 |
|
|
| 1011 |
|
if CN_c>0
|
| 1012 |
|
% variable threshold condition (slow)
|
| 1013 |
|
for t=1:reducedSegmentLength
|
| 1014 |
|
CNtimeSinceLastSpike=CNtimeSinceLastSpike-ANdt;
|
| 1015 |
|
s=CN_E>CN_Th & CNtimeSinceLastSpike<0 ;
|
| 1016 |
|
CNtimeSinceLastSpike(s)=0.0005; % 0.5 ms for sodium spike
|
| 1017 |
|
dE =(-CN_E/CN_tauM + ...
|
| 1018 |
|
CNcurrentInput(:,t)/CN_cap+(...
|
| 1019 |
|
CN_Gk/CN_cap).*(CN_Ek-CN_E))*ANdt;
|
| 1020 |
|
dGk=-CN_Gk*ANdt./tauGk + CN_b*s;
|
| 1021 |
|
dTh=-(CN_Th-CN_Th0)*ANdt/CN_tauTh + CN_c*s;
|
| 1022 |
|
CN_E=CN_E+dE;
|
| 1023 |
|
CN_Gk=CN_Gk+dGk;
|
| 1024 |
|
CN_Th=CN_Th+dTh;
|
| 1025 |
|
CNmembranePotential(:,t)=CN_E+s.*(CN_Eb-CN_E)+CN_Er;
|
| 1026 |
|
end
|
| 1027 |
|
else
|
| 1028 |
|
% static threshold (faster)
|
| 1029 |
|
E=zeros(1,reducedSegmentLength);
|
| 1030 |
|
Gk=zeros(1,reducedSegmentLength);
|
| 1031 |
|
ss=zeros(1,reducedSegmentLength);
|
| 1032 |
|
for t=1:reducedSegmentLength
|
| 1033 |
|
% time of previous spike moves back in time
|
| 1034 |
|
CNtimeSinceLastSpike=CNtimeSinceLastSpike-ANdt;
|
| 1035 |
|
% action potential if E>threshold
|
| 1036 |
|
% allow time for s to reset between events
|
| 1037 |
|
s=CN_E>CN_Th0 & CNtimeSinceLastSpike<0 ;
|
| 1038 |
|
ss(t)=s(1);
|
| 1039 |
|
CNtimeSinceLastSpike(s)=0.0005; % 0.5 ms for sodium spike
|
| 1040 |
|
dE = (-CN_E/CN_tauM + ...
|
| 1041 |
|
CNcurrentInput(:,t)/CN_cap +...
|
| 1042 |
|
(CN_Gk/CN_cap).*(CN_Ek-CN_E))*ANdt;
|
| 1043 |
|
dGk=-CN_Gk*ANdt./tauGk +CN_b*s;
|
| 1044 |
|
CN_E=CN_E+dE;
|
| 1045 |
|
CN_Gk=CN_Gk+dGk;
|
| 1046 |
|
E(t)=CN_E(1);
|
| 1047 |
|
Gk(t)=CN_Gk(1);
|
| 1048 |
|
% add spike to CN_E and add resting potential (-60 mV)
|
| 1049 |
|
CNmembranePotential(:,t)=CN_E +s.*(CN_Eb-CN_E)+CN_Er;
|
| 1050 |
|
end
|
| 1051 |
|
end
|
| 1052 |
|
% disp(['CN_E= ' num2str(sum(CN_E(1,:)))])
|
| 1053 |
|
% disp(['CN_Gk= ' num2str(sum(CN_Gk(1,:)))])
|
| 1054 |
|
% disp(['CNmembranePotential= ' num2str(sum(CNmembranePotential(1,:)))])
|
| 1055 |
|
% plot(CNmembranePotential(1,:))
|
| 1056 |
|
|
| 1057 |
|
|
| 1058 |
|
% extract spikes. A spike is a substantial upswing in voltage
|
| 1059 |
|
CN_spikes=CNmembranePotential> -0.02;
|
| 1060 |
|
% disp(['CNspikesbefore= ' num2str(sum(sum(CN_spikes)))])
|
| 1061 |
|
|
| 1062 |
|
% now remove any spike that is immediately followed by a spike
|
| 1063 |
|
% NB 'find' works on columns (whence the transposing)
|
| 1064 |
|
% for each spike put a zero in the next epoch
|
| 1065 |
|
CN_spikes=CN_spikes';
|
| 1066 |
|
idx=find(CN_spikes);
|
| 1067 |
|
idx=idx(1:end-1);
|
| 1068 |
|
CN_spikes(idx+1)=0;
|
| 1069 |
|
CN_spikes=CN_spikes';
|
| 1070 |
|
% disp(['CNspikes= ' num2str(sum(sum(CN_spikes)))])
|
| 1071 |
|
|
| 1072 |
|
% segment debugging
|
| 1073 |
|
% plotInstructions.figureNo=98;
|
| 1074 |
|
% plotInstructions.displaydt=ANdt;
|
| 1075 |
|
% plotInstructions.numPlots=1;
|
| 1076 |
|
% plotInstructions.subPlotNo=1;
|
| 1077 |
|
% UTIL_plotMatrix(CN_spikes, plotInstructions);
|
| 1078 |
|
|
| 1079 |
|
% and save it
|
| 1080 |
|
CNoutput(:, reducedSegmentPTR:shorterSegmentEndPTR)=...
|
| 1081 |
|
CN_spikes;
|
| 1082 |
|
|
| 1083 |
|
|
| 1084 |
|
%% IC ----------------------------------------------
|
| 1085 |
|
% MacGregor or some other second order neurons
|
| 1086 |
|
|
| 1087 |
|
% combine CN neurons in same channel,
|
| 1088 |
|
% i.e. same BF & same tauCa
|
| 1089 |
|
% to generate inputs to single IC unit
|
| 1090 |
|
channelNo=0;
|
| 1091 |
|
for idx=1:nCNneuronsPerChannel:nCNneurons-nCNneuronsPerChannel+1;
|
| 1092 |
|
channelNo=channelNo+1;
|
| 1093 |
|
CN_PSTH(channelNo,:)=...
|
| 1094 |
|
sum(CN_spikes(idx:idx+nCNneuronsPerChannel-1,:));
|
| 1095 |
|
end
|
| 1096 |
|
|
| 1097 |
|
[alphaRows alphaCols]=size(ICtrailingAlphas);
|
| 1098 |
|
for ICneuronNo=1:nICcells
|
| 1099 |
|
ICcurrentTemp(ICneuronNo,:)= ...
|
| 1100 |
|
conv2(CN_PSTH(ICneuronNo,:), IC_CNalphaFunction);
|
| 1101 |
|
% Changed conv to conv2 because it runs faster. (Andreas)
|
| 1102 |
|
end
|
| 1103 |
|
|
| 1104 |
|
% add the unused current from the previous convolution
|
| 1105 |
|
ICcurrentTemp(:,1:alphaCols)=ICcurrentTemp(:,1:alphaCols)...
|
| 1106 |
|
+ ICtrailingAlphas;
|
| 1107 |
|
% take what is required and keep the trailing part for next time
|
| 1108 |
|
inputCurrent=ICcurrentTemp(:, 1:reducedSegmentLength);
|
| 1109 |
|
ICtrailingAlphas=ICcurrentTemp(:, reducedSegmentLength+1:end);
|
| 1110 |
|
|
| 1111 |
|
if IC_c==0
|
| 1112 |
|
% faster computation when threshold is stable (C==0)
|
| 1113 |
|
for t=1:reducedSegmentLength
|
| 1114 |
|
s=IC_E>IC_Th0;
|
| 1115 |
|
dE = (-IC_E/IC_tauM + inputCurrent(:,t)/IC_cap +...
|
| 1116 |
|
(IC_Gk/IC_cap).*(IC_Ek-IC_E))*ANdt;
|
| 1117 |
|
dGk=-IC_Gk*ANdt/IC_tauGk +IC_b*s;
|
| 1118 |
|
IC_E=IC_E+dE;
|
| 1119 |
|
IC_Gk=IC_Gk+dGk;
|
| 1120 |
|
ICmembranePotential(:,t)=IC_E+s.*(IC_Eb-IC_E)+IC_Er;
|
| 1121 |
|
end
|
| 1122 |
|
else
|
| 1123 |
|
% threshold is changing (IC_c>0; e.g. bushy cell)
|
| 1124 |
|
for t=1:reducedSegmentLength
|
| 1125 |
|
dE = (-IC_E/IC_tauM + ...
|
| 1126 |
|
inputCurrent(:,t)/IC_cap + (IC_Gk/IC_cap)...
|
| 1127 |
|
.*(IC_Ek-IC_E))*ANdt;
|
| 1128 |
|
IC_E=IC_E+dE;
|
| 1129 |
|
s=IC_E>IC_Th;
|
| 1130 |
|
ICmembranePotential(:,t)=IC_E+s.*(IC_Eb-IC_E)+IC_Er;
|
| 1131 |
|
dGk=-IC_Gk*ANdt/IC_tauGk +IC_b*s;
|
| 1132 |
|
IC_Gk=IC_Gk+dGk;
|
| 1133 |
|
|
| 1134 |
|
% After a spike, the threshold is raised
|
| 1135 |
|
% otherwise it settles to its baseline
|
| 1136 |
|
dTh=-(IC_Th-Th0)*ANdt/IC_tauTh +s*IC_c;
|
| 1137 |
|
IC_Th=IC_Th+dTh;
|
| 1138 |
|
end
|
| 1139 |
|
end
|
| 1140 |
|
|
| 1141 |
|
ICspikes=ICmembranePotential> -0.01;
|
| 1142 |
|
% now remove any spike that is immediately followed by a spike
|
| 1143 |
|
% NB 'find' works on columns (whence the transposing)
|
| 1144 |
|
ICspikes=ICspikes';
|
| 1145 |
|
idx=find(ICspikes);
|
| 1146 |
|
idx=idx(1:end-1);
|
| 1147 |
|
ICspikes(idx+1)=0;
|
| 1148 |
|
ICspikes=ICspikes';
|
| 1149 |
|
|
| 1150 |
|
nCellsPerTau= nICcells/nANfiberTypes;
|
| 1151 |
|
firstCell=1;
|
| 1152 |
|
lastCell=nCellsPerTau;
|
| 1153 |
|
for tauCount=1:nANfiberTypes
|
| 1154 |
|
% separate rates according to fiber types
|
| 1155 |
|
% currently only the last segment is saved
|
| 1156 |
|
ICfiberTypeRates(tauCount, ...
|
| 1157 |
|
reducedSegmentPTR:shorterSegmentEndPTR)=...
|
| 1158 |
|
sum(ICspikes(firstCell:lastCell, :))...
|
| 1159 |
|
/(nCellsPerTau*ANdt);
|
| 1160 |
|
firstCell=firstCell+nCellsPerTau;
|
| 1161 |
|
lastCell=lastCell+nCellsPerTau;
|
| 1162 |
|
end
|
| 1163 |
|
|
| 1164 |
|
ICoutput(:,reducedSegmentPTR:shorterSegmentEndPTR)=ICspikes;
|
| 1165 |
|
|
| 1166 |
|
% store membrane output on original dt scale
|
| 1167 |
|
if nBFs==1 % single channel
|
| 1168 |
|
x= repmat(ICmembranePotential(1,:), ANspeedUpFactor,1);
|
| 1169 |
|
x= reshape(x,1,segmentLength);
|
| 1170 |
|
if nANfiberTypes>1 % save HSR and LSR
|
| 1171 |
|
y=repmat(ICmembranePotential(end,:),...
|
| 1172 |
|
ANspeedUpFactor,1);
|
| 1173 |
|
y= reshape(y,1,segmentLength);
|
| 1174 |
|
x=[x; y];
|
| 1175 |
|
end
|
| 1176 |
|
ICmembraneOutput(:, segmentStartPTR:segmentEndPTR)= x;
|
| 1177 |
|
end
|
| 1178 |
|
|
| 1179 |
|
% estimate efferent effects.
|
| 1180 |
|
% ARis based on LSR units. LSR channels are 1:nBF
|
| 1181 |
|
if nANfiberTypes>1 % AR is multi-channel only
|
| 1182 |
|
ARAttSeg=sum(ICspikes(1:nBFs,:),1)/ANdt;
|
| 1183 |
|
[ARAttSeg, ARboundary] = ...
|
| 1184 |
|
filter(ARfilt_b, ARfilt_a, ARAttSeg, ARboundary);
|
| 1185 |
|
ARAttSeg=ARAttSeg-ARrateThreshold;
|
| 1186 |
|
ARAttSeg(ARAttSeg<0)=0; % prevent negative strengths
|
| 1187 |
|
% scale up to dt from ANdt
|
| 1188 |
|
x= repmat(ARAttSeg, ANspeedUpFactor,1);
|
| 1189 |
|
x=reshape(x,1,segmentLength);
|
| 1190 |
|
ARattenuation(segmentStartPTR:segmentEndPTR)=...
|
| 1191 |
|
(1-ARrateToAttenuationFactor* x);
|
| 1192 |
|
ARattenuation(ARattenuation<0)=0.001;
|
| 1193 |
|
else
|
| 1194 |
|
% single channel model; disable AR
|
| 1195 |
|
ARattenuation(segmentStartPTR:segmentEndPTR)=...
|
| 1196 |
|
ones(1,segmentLength);
|
| 1197 |
|
end
|
| 1198 |
|
|
| 1199 |
|
% MOC attenuation using HSR response only
|
| 1200 |
|
% Separate MOC effect for each BF
|
| 1201 |
|
HSRbegins=nBFs*(nANfiberTypes-1)+1;
|
| 1202 |
|
rates=ICspikes(HSRbegins:end,:)/ANdt;
|
| 1203 |
|
for idx=1:nBFs
|
| 1204 |
|
[smoothedRates, MOCboundary{idx}] = ...
|
| 1205 |
|
filter(MOCfilt_b, MOCfilt_a, rates(idx,:), ...
|
| 1206 |
|
MOCboundary{idx});
|
| 1207 |
|
% spont 'rates' is zero for IC
|
| 1208 |
|
MOCattSegment(idx,:)=smoothedRates;
|
| 1209 |
|
% expand timescale back to model dt from ANdt
|
| 1210 |
|
x= repmat(MOCattSegment(idx,:), ANspeedUpFactor,1);
|
| 1211 |
|
x= reshape(x,1,segmentLength);
|
| 1212 |
|
MOCattenuation(idx,segmentStartPTR:segmentEndPTR)= ...
|
| 1213 |
|
(1- MOCrateToAttenuationFactor* x);
|
| 1214 |
|
end
|
| 1215 |
|
MOCattenuation(MOCattenuation<0)=0.04;
|
| 1216 |
|
% segment debugging
|
| 1217 |
|
% plotInstructions.figureNo=98;
|
| 1218 |
|
% plotInstructions.displaydt=ANdt;
|
| 1219 |
|
% plotInstructions.numPlots=1;
|
| 1220 |
|
% plotInstructions.subPlotNo=1;
|
| 1221 |
|
% UTIL_plotMatrix(ICspikes, plotInstructions);
|
| 1222 |
|
|
| 1223 |
|
end % AN_spikesOrProbability
|
| 1224 |
|
segmentStartPTR=segmentStartPTR+segmentLength;
|
| 1225 |
|
reducedSegmentPTR=reducedSegmentPTR+reducedSegmentLength;
|
| 1226 |
|
|
| 1227 |
|
|
| 1228 |
|
end % segment
|
| 1229 |
|
|
| 1230 |
|
disp('CONVOLUTION_CHANGE_TEST (if followed by zero all is good)')
|
| 1231 |
|
disp(max(CONVOLUTION_CHANGE_TEST)) %% for debugging
|
| 1232 |
|
|
| 1233 |
|
|
| 1234 |
|
%% apply refractory correction to spike probabilities
|
| 1235 |
|
|
| 1236 |
|
% switch AN_spikesOrProbability
|
| 1237 |
|
% case 'probability'
|
| 1238 |
|
% ANprobOutput=ANprobRateOutput*dt;
|
| 1239 |
|
% [r nEpochs]=size(ANprobOutput);
|
| 1240 |
|
% % find probability of no spikes in refractory period
|
| 1241 |
|
% pNoSpikesInRefrac=ones(size(ANprobOutput));
|
| 1242 |
|
% pSpike=zeros(size(ANprobOutput));
|
| 1243 |
|
% for epochNo=lengthAbsRefractoryP+2:nEpochs
|
| 1244 |
|
% pNoSpikesInRefrac(:,epochNo)=...
|
| 1245 |
|
% pNoSpikesInRefrac(:,epochNo-2)...
|
| 1246 |
|
% .*(1-pSpike(:,epochNo-1))...
|
| 1247 |
|
% ./(1-pSpike(:,epochNo-lengthAbsRefractoryP-1));
|
| 1248 |
|
% pSpike(:,epochNo)= ANprobOutput(:,epochNo)...
|
| 1249 |
|
% .*pNoSpikesInRefrac(:,epochNo);
|
| 1250 |
|
% end
|
| 1251 |
|
% ANprobRateOutput=pSpike/dt;
|
| 1252 |
|
% end
|
| 1253 |
|
|
| 1254 |
|
path(restorePath)
|