Transcript Modeling the auditory pathway - Computer Science

School of Industrial Engineering
Department of Computer Science
Purdue University
Modeling the impact of
Auditory Training
Research Advisor:
Presented By:
Prof. Aditya Mathur
Alok Bakshi
March 10, 2006
Auditory Neuroscience Lab
Northwestern University, Evanston
1
Research Objective
To construct and validate a model to
understand the effect of treatment on
children with learning disabilities and/or
auditory disorders
2
Objective of This Meeting


To present our understanding of the auditory
pathway and progress made towards the goal of
obtaining a validated computational model of the
auditory pathway.
To discuss possible approaches to the construction
and validation of a model of the auditory pathway.
3
Background


Children with learning problems are unable to
discriminate rapid acoustic changes in speech
It was observed that “auditory training” improves the
ability to discriminate and identify an unfamiliar sound
[Bradlow et al. 1999]

Can a computational model reproduce this
observation?
4
Methodology



Study physiology of Auditory System
Simulate the auditory pathway making new
models/using existing models of individual
components
Validate it against experimental results pertaining to
auditory systems
5
Methodology – Cont’d


Mimic experimental results of auditory processing
tasks on children with disabilities to gain insight
about the causes of malfunction
Experiment with the validated model to asses the
effect of treatments on children with
auditory/learning disabilities
6
Auditory System
From Ear to Auditory Cortex
Transforms sound waves into distinct patterns of neural
activity
Integrated with information from other sensory systems
to guide behavior and intra-species communication
7
Auditory Pathways
Ascending Auditory Pathway
Information from both the ears is carried to higher
centers
(focus on it in this presentation)
Descending Auditory Pathway
Brain influences the processing of information
8
Human Ear
http://www.owlnet.rice.edu/~psyc351/Images/Ear.jpg
9
Ascending Auditory Pathway
http://emsah.uq.edu.au/linguistics/ic310/Gif/audpath.gif
10
Brainstem Evoked Auditory Potential
What does the potential represent?
Ensemble behavior? At what points in the
pathway?
http://www.iurc.montp.inserm.fr/cric/audition/english/audiometry/ex_ptw/voies_potentiel.jpg
http://www.iurc.montp.inserm.fr/cric/audition/english/audiometry/ex_ptw/e_pea2_ok.gif
11
Auditory Qualities
Hearing Involves perception of
Loudness
Pitch
Timbre
Sound Localization
12
Place Coding
Different regions of the basilar membrane vibrate
differentially at different frequencies
Thus place of maximum displacement gives
topographical mapping of frequency (Tonotopy)
Conserved throughout the auditory system
13
Phase Locking
Hair Cells follow waveform of low frequency
sounds
Resultant phase locking provide temporal
information in the form of inter-aural time
differences
14
Auditory Neuron

Cell bodies in Spiral Ganglion

Send axons to Cochlear Nucleus

Two Types

Type I: Innervate Inner Hair Cell

Type II: Innervate Outer Hair Cell
15
Tuning Curve
Intensity
Threshold
Characteristic Frequency
Frequency
16
Cochlear Nucleus

Auditory Nerves connects almost exclusively to
Ipsilateral Cochlear Nucleus

Three divisions

Anteroventral Cochlear Nucleus (AVCN)

Posteroventral Cochlear Nucleus (PVCN)

Dorsal Cochlear Nucleus (DCN)
17
Cochlear Nucleus

Contains neurons of different response types

Breaks up sound into pieces of qualitatively different
aspects

Encode these aspects and send them to higher
centers for higher processing
18
Superior Olivary Complex (SOV)

Receives bilateral ascending input from Ventral
Cochlear Nucleus

Essential for Sound Localization

Four Divisions

Medial Superior Olivary Complex

Lateral Superior Olivary Complex

Medial Nucleus of the Trapezoid Body

Periolivary Nuclei
19
Medial Superior Olive



Uses inter-aural time difference as a cue for sound
localization
Receives excitatory inputs from both anteroventral
cochlear nucleus
Cells work as Coincidence Detectors responding
when both inputs arrive at the same time
20
Lateral Superior Olive


Uses inter-aural intensity difference as a cue for sound
localization
It receives

Excitatory input from Ipsilateral Cochlear nucleus

Inhibitory input from Contralateral Cochlear Nucleus
21
Inferior Colliculus


Thought to be have Auditory-Space Map
Neurons in auditory-space map responds best to
sound originating from a specific region of space
22
Modeling Perspectives

Stochastic versus Deterministic

Phenomenological versus Noumenal

Level of abstraction

Computationally tractable

Resemble the actual system
23
Modeling Option - I




Modeling of Individual Neuron
[Hodgkin-Huxley
model etc.]
Identification of anatomically different units/sub-units
in auditory pathway
Separate modeling of units by simulating many
neurons with appropriate parameters
Auditory pathway simulation by simulating these
units
24
Option – I Cont’d


Advantages

Nearer to reality

Easy to validate against experimental data
Disadvantage

Computationally intensive
25
Option – I Cont’d
Soma
Dendrites
Axons
Interneuron
Neuron Model
Auditory Pathway Unit
26
Option –I Cont’d
Input
Unit 1
Unit 2
Feedback ???
Unit 4
Unit 3
Output
27
Modeling Option - II



Identify functionally different units of auditory
pathway
Define and model input/output relationship
for these units
Simulate the auditory pathway by simulating
these units together
28
Option – II Cont’d

Advantages



Computationally tractable
Model gives more insight about the system
Disadvantage


Doesn’t represent biological reality completely
Don’t have complete understanding
29
Option – II Cont’d
Encode
Intensity
Sound
Encode
Frequency
Interpretation of Sound
Encode
Timbre
30
Neuron Models

Binary Neuron


[Olshausen
B. A. 2004 Sparse coding of sensory inputs]
On/off depending on the input
Firing Rate Neuron
[Tanaka
S. 2001 Computational approaches to the architecture and
operations of the prefrontal cortical circuit for working memory. ]


Firing rate instead of individual spikes are modeled
Integrate and Fire model
[Izak,
R. 1999 Sound source localization with an
integrate-and-fire neural system ]

Hodgkin-Huxley model
[Hodgkin
A. et. al. 1952 Measurement of current-voltage
relations in the membrane of the giant axon of Loligo]
31
Neuron Models – Cont’d


Hodgkin-Huxley model
 Chaotic but completely deterministic
Approximation Algorithm Fox R. F. 1997 Stochastic Versions of the Hodgkin[
Huxley Equations]
White noise term in HH model
Channel State Tracking Algorithm


[Rubinstein
1995 Threshold
Fluctuations in an N Sodium Channel Model of the Node of Ranvier ]
Simple but computationally intensive
Channel Number Tracking Algorithm Gillespie D. T. 1977 Exact


[
Stochastic Simulation of Coupled Chemical Reactions]

Computationally efficient
32
Molecular basis
-70mV
Na+
action
potential
K+
Ca2+
Ions/proteins
Gerstner W. and Kistler W., Spiking Neuron Models’0233
Action Potential
Voltage
Spike
Time
Hyper-Polarization
34
Ion Channels

Each channel opens with rate ai and closes with rate bi

Potassium ion channel


Has four similar sub-units

Each subunit is open or closed independently

Open iff all four sub-units are open
Sodium Ion channel

Three similar sub-units and one slow sub-unit

The channel id open iff all four sub-units are open
35
Channel Kinetics
www.sis.ipm.ac.ir/seminars/weekly%20seminars/course/Neural%20modeling/babadi04.ppt
36
Binary Neuron



Each neuron has two states

On (1)

Off (0)
Each input to the neuron has a particular weight-age
If the combined input exceeds threshold then neuron
comes into on (1) state
37
Firing Rate Neuron



The firing rate is a function of voltage
Firing rate rather than individual spikes are
modeled
Hence encodes information related with firing
rate and ignores spikes
38
Integrate and Fire Neuron




Time of occurrence of Action Potential is modeled rather than its
shape
Dynamics of Neuron

Sub-Threshold

Supra-Threshold
Conductance due to Na and K channels ignored in SubThreshold voltage
If voltage becomes greater than threshold

A spike is generated

Membrane potential is reset to a value for refractory period
39
Hodgkin-Huxley Model
m
m
t
m
m
40
Hodgkin-Huxley Model –Cont’d
ai and bi


Functions of voltage V
Hodgkin-Huxley model successfully describes
the mechanism of Action Potential
The model is completely deterministic
41
Stochastic Phenomena

Kinetics of ion channels as continuous time discrete
state Markov jumping process

Channel noise affects

Stability of resting potential

Temporal representation of sound
42
Ion Channel Kinetics for Na
Mino H. et al. Comparison of Algorithms for the Simulation of
Action Potentials with Stochastic Sodium Channels’ 02
43
Approximation Algorithm

Langevin description of cellular automaton model

Channel density variable instead of modeling
individual ion channels

Computationally less intensive but poor performance
44
Exact Algorithms


Channel State Tracking Algorithm

Tracks state of each individual channel

Simple but more computation requirement
Channel Number Tracking Algorithm

Tracks number of channel in each state

Assumes multiple channels are memory-less

Computationally quite efficient
45
Validation

Validate against what?


Auditory Evoked Responses
Data from other animals ???
46
Progress so far…


Studied anatomical structure of the auditory
pathway
Surveyed various models of neuron and
neural networks
47
References
•
•
•
•
•
Drawing/image/animation from "Promenade around the
cochlea" <www.cochlea.org> EDU website by R. Pujol et al.,
INSERM and University Montpellier
Fox F. R. 1997, Stochastic versions of the Hodgkin-Huxley
Equations. Biophysical Journal, Volume 72, 2068-2074
Gunter E. and Raymond R. , The central Auditory System’ 1997
Kraus N. et. al, 1996 Auditory Neurophysiologic Responses and
Discrimination Deficits in Children with Learning Problems.
Science Vol. 273. no. 5277, pp. 971 – 973
Mino H. et al. 2002, Comparison of Algorithms for the
Simulation of Action Potentials with Stochastic Sodium
Channels. Annals of Biomedical Engineering, Vol. 30, pp. 578587
48
References – Cont’d
Purves et al, Neuroscience 3rd edition
•P. O. James, An introduction to physiology of hearing 2nd
edition
•Ruggero M. A. and Rich N. C. 1991, Furosemide alters Organ
of Corti mechanics: Evidence for feedback of Outer Hair Cells
upon the Basilar Membrane. The Journal of Neuroscience,
11(4): 1057-1067
•Tremblay K., 1997 Central auditory system plasticity:
generalization to novel stimuli following listening training. J
Acoust Soc Am. 102(6):3762-73.
•
49