KEELE COCHLEA/COCHLEAR NERVE MODEL
Professor E.F. Evans, Department of Communication and Neuroscience, Keele University, Keele, Staffs, ST5 5BG.
Tel: 01394 388295; email:
This software model simulates cochlea and cochlear nerve response properties to stimuli which are under full control of the operator. It runs as a stand-alone LabVIEW Application under Microsoft Windows (including XP).
1. Installation
1. Create \LabView directory.
2. Copy "model10s.exe" to \LabView directory.
2. To start the program.
In Windows, run model10s.exe from the \ LabView directory
The programme will start, filling the screen with the "control panel" of the model (see Figure attached), and generating automatically a shaped tone at 1 kHz, exciting a cochlear nerve fibre with a characteristic frequency (CF) of 1 kHz at 40dB sound level (about 40 dB above threshold). The tone will be heard over a loudspeaker plugged into the audio output of the PC, and also the spike discharges from the cochlear nerve fibre as clicks. The action potential spikes are visible in the yellow graphic panel in the middle of the screen.
3. Navigating the Model.
Each section has its own local help indicated by the red button. Clicking on the red button will produce a help window and halt the model until you click on the "OK" button in the help window. To obtain an overall description of the model and overall help, click on the square red MAIN HELP button towards the right middle of the control panel (scrolling the screen right if necessary to see it - see 3a above.) (The help windows can be dragged around by holding down the left mouse button on the blue border).
4. To Use the Model
General
The model allows you to have complete control over the frequency and level, delay, duration and rise-fall-time of a tonal stimulus, with or without noise masking, and to examine the effects of the stimulus on the spike discharges of a single fibre in the mammalian cochlear nerve. You can hear the sound over the audio loudspeaker by clicking on the Signal Sound button (clicking again toggles the sound off). You can select the CF (characteristic frequency - the frequency to which the nerve fibre responds best, related to its position of origin along the cochlea), and the degree of tuning, of the cochlear fibre.
The model can be stopped by clicking on the Stop button on the top border of the window. Restart the model by clicking on the Start arrow (->) further to the left.
The results are displayed in a number of ways:-
The Tone Signal shows the waveform of the whole stimulus; the Tone Waveform, a magnified view of the onset to show the effects of changing the rise-fall time.
The amplitude and phase characteristics of the cochlear filter connected to the cochlear nerve fibre are represented as the "Impulse Response" (ie, the response of the cochlear filter to a click stimulus).
The Filtered O/P (Output) shows the results of filtering the chosen stimulus by the chosen cochlear filter. It is an approximation to the intracellular inner hair cell receptor potential.
The Raw Spikes O/P are seen in an oscilloscope type display. Clicking on the Spike Sound button allows you to hear them as clicks over the PC's loudspeaker (clicking again toggles the sound off).
The Raw Gated Spikes meter display shows the instantaneous spike count during the count "gate" shown under the raw spike record.
The PSTH (Peri-Stimulus Time Histogram) shows the average probability of response as a function of time before, during and after the stimulus.
The PH (Period Histogram) shows the average probability of discharge corresponding to a single period of the stimulus.
The Gated Spike CountChart shows a record of the running spike counts. Dragging the centre of the button below the chart allows you to review the chart's past history.
Tour of Features
1. Start by experimenting with the effect of changing the Tone Frequency:-
Drag the yellow pointer around the dial by holding the left mouse button down while dragging the pointer. The frequency obtained is indicated under the main dial. Adjust as desired with the fine control.
Clicking on the left white disk button will continuously increment the frequency from the dial position (the white disk on the right will decrement it). Clicking on the white disk button again, resets the frequency to the dial setting.
Clicking the mouse on the Signal Sound button allows you to hear the stimulus over the audio speaker. Listen to how the sound changes while you change the frequency of the tone signal, and observe the effect on the signal waveform as shown in the Tone Signal and Waveform displays. The Tone Signal display shows the entire tone signal waveform on a millisecond timebase, and the Tone Waveform shows a magnified view of the first 30 ms or so of the tone onset, in order that the effect of changing the Rise-Fall time (see below) can be more clearly seen.
2. Tone Level:-
This slider sets the intensity of the tone (not heard as changes in loudness in the loudspeaker).
Move the pointer with the mouse (holding down left mouse button), or click on the up or down arrows of the slider value control.
Clicking on the white disk buttons will continuously increment (lower disk) or decrement (higher disk) the level from that shown on the slider control. The actual tone level is shown on the indicator below the slider.
3. Tone Onset, Duration, and Rise-Fall Times:-
These control the onset delay, duration and rise-fall times of the tone stimulus. The default values are 10, 50, and 5 ms respectively.
Change each value by clicking on its up or down arrows.
Observe the effect of changing the rise-fall time to 0 on the Filtered Output (the imaginary output of the cochlear filter), particularly when the Tone Frequency is off-CF. For example, with the CF set at 1 kHz, and the Tone Frequency set at 0.4 kHz, the onset and offset transients will be clearly seen in the Filtered Output. Changing the rise-fall time back to 5 ms will progressively minimize the unwanted transients, which represent "splatter" of the signal from the intended frequency to neighbouring, unwanted frequencies.
4. Cochlear Nerve Filter (ie as established by the active mechanisms of the Organ of Corti):-
Change the CF (Characteristic Frequency - the frequency to which the cochlear nerve fibre responds best) by dragging the slider along with the mouse (holding down the left mouse button while dragging). Letting go the mouse button leaves the slider in the current position. This is equivalent to choosing a different cochlear nerve fibre along the tonotopic array of nerve fibres emanating from the cochlear partition.
The CF chosen is indicated by the value above the slider, and this can be editted if required directly, or by clicking on the up and down arrows. Try changing the CF to 2 kHz. The evoked response will disappear. Changing the Tone Frequency to 2 kHz will restore the response.
The sharpness of tuning of the cochlear filter is controlled by typing in new values after clicking on the existing value with the mouse. A value of 1 represents the normal sharpness of tuning. A value of 0.2 deteriorates the tuning by a factor of 5.
Reset Tone Frequency, CF, and Sharpness of Tuning to 1.00.
5. Impulse Response:-
This represents the amplitude and phase characteristics of the cochlear filter (connected to the cochlear nerve fibre) in terms of its "impulse response". This is the response of the filter to an impulse (click stimulus). Note how the impulse response changes with changing the CF and Sharpness ofTuning of the cochlear filter.
6. Filtered Output:-
This is the result of filtering your chosen signal with the chosen cochlear filter (indicated by its impulse response). If masking noise is present, you will see this also on the silent baseline.
The filtered output is roughly equivalent to the inner hair cell receptor potential at low levels. (At high levels and frequencies, the dc receptor potential would predominate).
7. Raw Spikes Output:-
This shows the action potential "spikes" generated by the cochlear nerve fibre in response to the stimuli, on an oscilloscope-type display on the same time scale as the Tone Signal above.
You can hear the spikes as clicks over the audio speaker by clicking on the Spike Sound button. Clicking on it again turns the sound off.
8. Gated Spikes:-
The cochlear fibre spikes are counted during a count "gate" indicated under the raw spike record by the raised Spike Gate whose onset and duration can be set as desired to encompass the stimulus (default 10 and 60 ms respectively).
The Raw (stimulus by stimulus) Gated Spike count is displayed on the meter display with the values indicated immediately beneath.
The Mean Gated Spike count, averaged since the last START (->) is indicated below and to the left. (The Mean Sp/sec is the spike rate in spikes per sec averaged over the whole duration of the stimulus (0 to 130 ms). The number of repetitions of the signal in these averages is given lower down in the display.
9. Spike Gate:-
This controls the time during which spikes are counted for the gated spike displays.
Change the onset and duration of the gate from the default values (10, 60 ms) by clicking on the up/down arrows associated with each value control.
10. Tone/Noise buttons:-
Clicking on either or both of these buttons allows you to select tone, noise or a mixture of both signals. To turn one or other off, click on the respective button again.
The noise signal is continuous (though not heard as such in the PC loudspeaker). Its level can be set by clicking on the up or down arrows of its level control, or by typing in the required value. A value of about 20 dB will reach threshold for the cochlear fibre response. Try increasing the noise level until the tone response is masked out. Resetting the PSTH, by clicking on STOP and the START arrow (->), will allow you to see if the rate response has been masked out. Note that the PH continues to show phase locking to the tone 10 dB or so above the noise level required to mask the rate response as seen in the PSTH.
11. PSTH (Peri-Stimulus Time Histogram):-
Shows the probability of discharge before, during and after the tone stimulus along the total time duration of the stimulus signal (120 ms). The probability is averaged over the number of repetitions of the stimulus (indicated to the right of the PSTH). The bin width (PSTH: No of Samples per bin) of the histogram is adjusted by the control to the right of the PSTH display. Try increasing to 20. This gives a more coarse resolution, but a more quickly convergent PSTH will result. Smaller values than 10 will give more detailed resolution, but will take longer for a clear shape for the PSTH to be seen. Reset to the default, 10 (= 1 ms per bin).
The PSTH is reset by stopping and Starting (->) the model with the buttons at the top of the screen window.
12. Period Histogram:-
This displays the probability of discharge as a function of one period (wavelength) of the stimulus waveform. Thus, for a 1 kHz stimulus tone, the histogram covers 1 ms (one period or cycle; 10 samples per period at 10 kHz sampling rate). To ensure that this is the case, set the No of samples per period (lower control) to the same value as the upper indicator (No of samples per period to set).
Phase-locking of the spikes to the stimulus cycle shows as a peak occupying about half of the total period. The peak may be "wrapped round" the ends of the display window.
The PH is reset by stopping and Starting (->) the model with the buttons at the top of the screen window. This needs to be done each time after changing the No of samples per period.
Scroll the screen down to see phase-locking demonstrated with the spike and waveform displays at the bottom of the model (see below).
13. Gated Spike Count Chart:-
This displays the running gated spike counts over a long duration. The past history of the chart can be retrieved by scrolling the middle button of the slider below the chart by holding the left mouse button down whilst dragging the slider button with the mouse.
14. Phase-locking:-
The upper trace shows a magnified view of the Raw Spikes Output and of the tone waveform, during the tone signal (20 to 30 ms along).
Change the CF and the Tone Frequency to 0.2 kHz (Tone level 40 dB). Note that the spikes tend to be triggered near the peak of the waveform each time they occur, with a small time 'jitter'. This is 'phase-locking' of the spikes to a specific phase (position along the cycle) of the waveform. The time intervals between the spikes are at multiples of the period of the tone stimulus (one, two, three periods, etc.). They change with the tone frequency. Try changing this to 0.4 kHz (together with the CF). Note how the intervals shorten. To see the phase-locking quantitatively with the CF and Tone Frequency at 0.400 kHz, change the Period Histogram (PH) No of samples to 25, and restart the model. Note a (half-wave rectified) version of the waveform building up in the PH (bins 0 to 12).
Phase-locking persists up to tone frequencies of about 5 kHz.
Reset the model to 1 kHz for CF and Tone Frequency; No of samples for the PH to 10.
5. Appendix: Demonstration of Model at Keele Meeting of Physiological Society, April 20th, 1995
A computational model of the cochlea and cochlear nerve for teaching and research.
E.F. Evans, Department of Communication & Neuroscience, Keele University, Keele, Staffordshire ST5 5BG.
Since 1974, we have had in routine use in the laboratory, a hardware analogue of the cochlea and cochlear nerve (Evans, 1979, 1980). This simulates, in real-time, the salient features of the mammalian single cochlear nerve fibre discharges. It has proved valuable for three reasons. First, it has allowed rehearsal and optimization of manual and computer-controlled strategies of response analyses, particularly automated threshold and receptive field mapping (e.g. Evans, 1979, Evans, et al, 1987). Second, it has proved valuable for teaching purposes, enabling students to have `hands-on' experience of determining thresholds, receptive fields and rate-level functions, phase-locking and other suprathreshold properties of cochlear nerve fibres. Thirdly, it, and its software successor, is allowing a systematic investigation of the importance of certain features of the model in determining the matching of its performance to the responses of mammalian nerve fibres to a wide range of complex stimuli (eg Evans, 1987, 1988, 1989).
Unlike the hardware model, this version can be run in its entirety on a PC, complete with the generation of signal inputs and visual and auditory output. The computational model takes the above hardware model as its basis. It is constructed in LabVIEW (National Instruments U.K. Corp., 21 Kingfisher Court, Hambridge Road, Newbury, Berks, RG14 5SJ) and can run somewhat faster than real-time on a 66MHz 486DX PC. It accepts stimuli selected by mouse or keyboard from a repertoire of previously synthesized stimuli or generated in real-time as directed by `controls' displayed on the monitor and manipulated by mouse or keyboard. It generates an oscilloscope-like view of the output action potentials in time, audible as clicks over the PC's loudspeaker or sound card. For research purposes, the parameters of the model can be easily manipulated via an array of `front-panel controls' on the display, and the responses of any stage of the model visualized by an oscilloscope-like display, or written to file. It reproduces over a wide range of stimulus levels (100dB) the following characteristics: suprathreshold tuning properties to tones and noise, discharge rates, intensity functions, post-stimulus time, period and inter-spike interval histograms, autocorrelograms in response to tones, noise, clicks and multi-component stimuli and reverse correlograms to noise.
LabVIEW has the singular advantages that the model's development, modification, testing and display of output can be accomplished very rapidly without writing any text code, by the graphical `wiring up' on screen of iconized functions from a very large function library. New functions can be built up from the existing functions and/or from purpose-built functions constructed from simple formulae or `C' code. Conventional programming facilities such as file and array handling, while- and for-loops, case, and sequence structures, etc., are easily incorporated. A standalone, compiled version of the model can be generated for tamper-free class use.
Physiological justification for the stages of the model is summarized in Evans (1975). Briefly, a front-end bandpass filter filters the incoming signal by convolution with a chosen impulse response (gammatone or reverse correlogram taken from our physiological data). An AGC limiter simulates saturation of the mean discharge rate without corrupting the period histogram of stimuli at levels above saturation. A halfwave rectifier simulates unidirectional excitation at the hair cell level. A logarithmic compression stage simulates the relation between spike rate and stimulus amplitude between threshold and saturation. High pass filtering provides adaptation and off-suppression. Probabilistic spike generation is simulated by two-quadrant multiplication between the processed signal and noise followed by level detection, so that the probability of discharge is related to the halfwave rectified stimulus waveform. An absolute refractory period of 2ms follows each spike.