Auditory nerve response model (Zhang et al 2001)

 Download zip file 
Help downloading and running models
A phenomenological model was developed to describe responses of high-spontaneous-rate auditory-nerve (AN) fibers, including several nonlinear response properties. The implementation of this model represents a relatively simple phenomenological description of a single mechanism that underlies several important nonlinear response properties of AN fibers. The model provides a tool for studying the roles of these nonlinearities in the encoding of simple and complex sounds in the responses of populations of AN fibers.
1 . Zhang X, Heinz MG, Bruce IC, Carney LH (2001) A phenomenological model for the responses of auditory-nerve fibers: I. Nonlinear tuning with compression and suppression. J Acoust Soc Am 109:648-70 [PubMed]
Citations  Citation Browser
Model Information (Click on a link to find other models with that property)
Model Type:
Brain Region(s)/Organism:
Cell Type(s): Cochlea hair outer GLU cell; Auditory nerve;
Gap Junctions:
Simulation Environment: C or C++ program;
Model Concept(s): Temporal Pattern Generation;
Implementer(s): Zhang, Xuedong ;
Search NeuronDB for information about:  Cochlea hair outer GLU cell;
/** \mainpage AN model 2 Documentation for CAT\n source file navigation
 *\section download Download and uncompress the source files
 *<p> Click here to download the source file <a href="">ancat.tar.gz</a>
 *<p> Uncompress it using the command <I>tar -xzvf ancat.tar.gz</I>\n
 * or <I>gunzip -dc ancat.tar.gz | tar -xv</I>.
 *\section compile Compile the file
 * <p>In the source file directory type\n <I>make anmodel</I>
 * <p>If you don't have g++, try c++ instead( and change the CC value to c++ in <I>makefile</I>).
 *\section run Run the model
 * <p>The program <I>anmodel</I> is used if you want to pass a stimulus through the AN filter bank.\n
 * The typical usage of the program is as follows:\n
 * <I>./anmodel -wavefile click -reptim 0.020 -fibers 32 -cflo 200 -cfhi 20000</I>
 * <p>Parameter values are set by specifying several options, each of which is preceded by a "-". In
 * this example the <I>-wavefile</I> option is set to file name <I>click</I>, which holds an ascii
 * stimulus file(single column of values.) The sampling rate of the stimulus should match that used in
 * the simulation; the default values is 500kHz but this value can be changed with the option
 * <I>-tdres</I>.\n
 * <p>--The <I>-reptim</I> option determines how long(in seconds) the model should run, it will append
 * zero if the file input is shorter than the value of reptim.\n
 * The <I>-fibers</I> option determines how many filters(fibers) you want to include in the filter
 * bank(the Maximum is 64).\n
 * <p>--The CFs of the fiber in the filterbank can be set using the <I>-cflo</I> and <I>-cfhi</I> options
 * (as shown), or by setting the cf of the fiber in the center of the bank(<I>-cf</I>) and setting
 * the distance between fibers(<I>-delx</I>) in terms of mm along the BM(using Liberman's CF map). If
 * <I>-cflo</I> and <I>-cfhi</I> are used, the fibers are evenly apaced along the BM using Liberman's
 * map.
 * <p>To see all of the option and the default values, type <I>anmodel -help</I>.
 * <p>The output of the program is stored in file <I>filterout</I> and <I>synapseout</I>. Each column
 * of the data file is the output of one fiber fiter. The <I>filterout</I> stores the output of the
 * tuning filter(in Pascal) and <I>synapseout</I> stores the output of the synapse(in spikes/sec).
 * \section filelist Description of the files
 * \ref filelist_detail "go..."

 * \section FAQ Questions about the code
 * \ref faq_detail "go..."
 * \anchor faq_detail
 * <p>1.<I>How the ihc-ppi nonlinearity is determined?</I>\ref answer1 "answer"
 * <p>2.<I>What is the relationship between vsat and kppi?</I>\ref answer2 "answer"
 * <p>3.<I>How is the structure of the model defined in the code?</I>\ref model_structure "answer"
 * <p>4.<I>How are the parameters of the model set up?</I>\ref model_parameters "answer"
 * <p>5.<I>How the model processes the signal?</I>\ref model_run "answer"
 * <p>6.<I>How to set the model to right spontaneous rate and saturating rate?</I>
 *      \ref model_spont_faq "answer"
 *\section model_structure Model Structure
 *\ref model_structure_detail "go..."
 * \anchor model_structure_detail
 * class TAuditoryNerve defines the main structure of the AN model\n
 * class TBasilarMembrane defines the main structure of the tuning filter with control path
 *\section model_parameters How the parameters are set up
 * The parameters of the model are determined in the initialization of the model fiber\n
 * TAuditoryNerve::construct()\n
 * TAuditoryNerve::init()
 *\section model_run Run the model
 * This section tries to help you understand how the signal passes through the model
 * to get the final output.(Click on these function names to see the detail codes)\n
 * TAuditoryNerve::run() shows the main function that calls all other functions.\n
 * TBasilarMembrane::run() shows how the signal passes through the control path and time-varying
 * filter, and also shows how the nonlinear filter is changed instantaneously.\n
 * To see how the signal runs through a specific part of the model, click on these functions:\n
 * TGammaTone::run, TAfterOHC::run,THairCell::run, TSynapse_WS::run, TSpikeGenerator::run
 *\section other_answers Other answers
 * \anchor answer1 Answer1
 * <p>ihc-ppi nonlinearity is setup in TAuditoryNerve::init() :\n
 *   calculate kppi(kslope) as the parameter of the IHC-PPI nonliear function\n
 *   call TSynapse_WS::setppi_mgheinz(double) to set mgheinz's soft rectifier nonlinear function\n
 *   call TSynapse_WS::init_WS()(which call TIHCPPI_mgheinz::init() to calculate the real parameters
 * <p>Run\n
 *   TSynapse_WS::run call TIHCPPI_mgheinz::run to get the output of the nonlinear function
 *  The ihc-ppi nonlinear function used in the code
 *  is soft rectifier nonlinear function used by Michael G. Heinz(ref: Heinz et al. 2000) before.
 *  See TIHCPPI_mgheinz::run() for detail information about this nonlinear function
 * The parameters in this nonlinear function are basically determined by threshold(kppi/kslope) of the
 *  fiber, which is calculated in function TAuditoryNerve::init(), this function also calls
 *  the function TSynapse_WS::setppi_mgheinz(double) to set the nonlinear function as mgheinz's
 *  TIHCPPI_mgheinz::init() calculates the real parameters according to the kppi(kslope), which is called
 *  when the synapse is initialized(TSynapse_WS::init_WS()) in TAuditoryNerve::init()
 * \anchor answer2 Answer2
 * <p> the vsat is not used if the kppi is set(>0), see TIHCPPI_mgheinz::init
 * \anchor model_spont_faq Answer about spontanous rate
 * <p> The user can use the -spont to set the spontaneous rate of the fiber output
 * this value is estimated based on syanpse output(sout) without refractoriness, to set the spont
 * rate to the right value for spikes, you should set a little bit higher than desired. Say if you
 * set the spontaneous rate at 50 spikes/second in the model, the real spontaneous rate is around
 * 42 spikes/second
 * The saturation rate of the fiber is calculated based on the real spikes(around 250 spikes/sec), if
 * you using sout to calculate the sustained rate, you may get the rate higher than this one.
 * \anchor filelist_detail FileList\n
complex.cpp : routines about the complex number operation\n

filters.cpp : defines all the filters, including\n
	      1. gamma-tone filters including basilar membrane filtering & wideband filter in control path\n
              2. low pass filter in the hair cell\n
synapse.hpp : define the structure of the synapse, including\n
              1. ihc-ppi mapping\n
              2. neural transmitter release model(Westerman and Smith)\n
              3. spike generator including refractoriness\n
hc.cpp      : defines the haircell model(inner and outer), including\n
              1. nonlinear(arctan, and Boltzman) function\n
              2. haircell (consists of the nonlinearity and the low-pass filter)\n
cmpa.cpp    : defines the structure of the model, including\n
              1. basilar membrane (gamma-tone filter with forward control path)\n
              2. auditory nerve fiber model (basilar membrane, inner hair cell, synapse, spike generator)\n
	      3. other functions that can be used\n
anmodel.cpp :\n
             wrap up all the functions to do the appropriate test\n


/**\anchor init_synapse ********************* Init Synapse ********************************
  * for linear,mgheinz version, if we set vsat = 10*PImax(which is determined by kppi now),
  * that means when ihc comes to the maximum value(1) the ppi = vsat
  * while the  high spont fibers saturate at PImax, which is 20dB less than the
  * vsat, for low spont rate, this seems not true, so the low spont fiber will not
  * saturate at the same level as that of high spont fiber

  * The threshold of fibers seems also change with the spont rate, with the same vsat
  * the threshold of low spont fibers seems lower than that of the high spont fiber
  * which means that we should derease the vsat for low spont such that it has higher threshold
  * once we set the saturate level of IHC, we can adjust this value to
  * make the AN fibers saturate/threshold at desired level
  * for low spont, we set the vsat = PImax, such that the fibers will
  * saturates until ihc saturates
  * if so, why the threshold of low spont fiber will decrease?

 * For How the synapse are initialized see \ref init_synapse "init_synapse"

  change the order of wideband-filter, change the initial bandwidth of the filter,
  change the normalized gain\n

  Try get the wideband filter timevaring????

  difference from mgheinz's human(sout) and cat(spikes)
  cutoff = 400/800hz
  ass = 130/250

  try to using a smoothed tau: changes in the AfterOHC nonlinearity

   Now trying to work on the IHC nonliearity to get rid of the peak in the sync-level funcion
   Also trying to get the symetrical value by only including three time variable gammatone filter
   now : adjust afterohc NL parameters  : dc = dc+0.1
   kppi : ihc threshold is changed a little bit

   1/12: 1. decrease the Ass, when Ass = 250, the sustained rate is 400 for 50ms, 300 for 200ms
            This is too high compared with the data,(ref:Schalk,Sachs 1980)
         2. compare the noise/tone rate level function at high CF
            (ref:Schalk,Sachs 1980,

   1/5 : 1. using 5ms rise/fall(changes made in model.hpp,stim.cpp)
         2. using realspikes to caculate the rate
         3. when caculate the spikes, simulate the synapse state as in
            quite to include the refratoryness
	 4. adaptation time constant is the same as Ian's (2ms, 60ms)

   12/30 : more changes should be done:
   1. the fit of Q10 using Ian's equations
   2. using soft rectify between Vihc/kppi???
   3. The time constant is adjusted so the reasonable durations can be used
   4. better using spikes generator

   12/22 : change some parameters as said in Ian's email
   time constant : 2ms, 60ms
   Ass decreased to 150
   Linear between Vihc/kppi

   8/8/99 Now we try to clear up the code

  6/1/99 Now the wideband filter in control path is more flexible; the center frequency of the filter is shifted a little higher than cf,
  the gain of the filter is normalized to 1 at the frequency of cf;
  in structure T_Anpara: several parameters of wideband filter are added to satisfy such modification
  they (wb_deltaphase,wb_gain) are initialized in Init_Wideband

   add a high path tempory : an

/** \file MAINHELP
\brief Quick Reference

<p>in cmpa.cpp

// initialize the auditory nerve fibers and set the parameters

Object Description:
<p>In C++, all the functions and corresponding structures are wraped up together as classes(objects), An object
can have other objects and define the functions to manipulate them.
The main objects(classes) defined in the codes are followings:

class TAuditoryNerve  /*(cmpa.hpp,cmpa.cpp, same as T_Anpara)
  the main structure of the AN2 model :

   including four sub-structure(objects)
     a nonlinear basilar membrane filters : TBasilarMembrane (in cmpa.hpp)
     inner hair cell model                : THairCell        (in hc.hpp)
     synapse models                       : TSynapse_WS      (in synapse.hpp)
     spike generator                      : TSpikeGenerator  (in synapse.hpp)

   including functions including
     construct(void): create all the objects
     init(species, cf, spont) : set up all the parameters of the model
     double run(x) : run the stimuli step by step

class TBasilarMembrane  //(cmpa.hpp,cmpa.cpp)
  the main structure of the nonlinear basilar membrane :

   including four sub-structure(objects)
     a 4th-order gamma-tone filter          : TGammaTone       (in filters.hpp)
     wideband-pass filter in control path : TGammaTone       (in filters.hpp)
     outer hair cell model                : THairCell        (in hc.hpp)
     nonlinear after ohc                  : TNL_AfterOHC     (in hc.hpp)

   including functions including
     double run(x) : run the stimuli step by step

for other objects, please refer to the corresponding hpp file

File Description:
complex.cpp : routines about the complex number operation

filters.cpp : define all the filters, including
	      1. gamma-tone filter as basilar membrane filtering
              2. low pass filter in the hair cell

synapse.cpp : define the structure of the synapse, including
              1. ihc-ppi mapping
              2. neural transmitter release model(westman)
              3. spike generator including refractory

hc.cpp      : define the haircell model(inner and outer), including
              1. nonlinear(arctan, and boltzman) function
              2. haircell (consists of the nonlinear and the low-pass filter)

cmpa.cpp    : define the structure of the model, including
              1. basilar membrane (gamma-tone filter with forward control path)
              2. auditory nerve fiber model (basilar membrane, inner hair cell, synapse, spike generator)
	      3. other functions that can be used

anmodel.cpp :
             wrap up all the functions to do the appropriate test