University College London wwwlesicalabcom THE STRUCTURE AND FUNCTION OF THE AUDITORY SYSTEM HEARING THE PROBLEM Potential danger Sourc e of interest Competing sources Background noise Key points ID: 911147
Download Presentation The PPT/PDF document "Nicholas A. Lesica Ear Institute" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Nicholas A.
LesicaEar InstituteUniversity College Londonwww.lesicalab.com
THE STRUCTURE AND FUNCTION OF THE AUDITORY SYSTEM
Slide2HEARING
THE PROBLEM
Potential dangerSource of interest
Competing sources
Background noise
Key points:Hearing is a 360° senseAuditory objects are transparent(ish)
Slide3The ear receives, filters, and encodes acoustic information …
… and the brain analyzes it to update its internal model of the world
HEARINGTHE SOLUTION
Slide4Frequency
IntensityReceptive fieldTemporal profile
TimeActivity
sound
Low-D basis
Encoding
Ear
Feature extraction
Brainstem
Nonlinear recombination
MidbrainHigher cortexPerception
Selective amplification
Thalamocortical loop
High-D basis
High SNR basis
Classification
help
hello
yellow
mellow
Melo
Jello
fellow
THE MAMALIAN AUDITORY PATHWAY
OVERVIEW
Slide5THE EAR
OVERVIEW
Slide6THE OUTER EAR
The outer ear collects sound and amplifies low frequencies …
-15
°
0
°
+15°
… and also filters high frequencies in a location-dependent manner
Slide7THE EAR
OVERVIEW
Slide8THE MIDDLE EAR
The middle ear compensates for the impedance mismatch between the air in the ear canal and the fluid in the cochlea …
e
ar drum
E
ar canal (air)
Cochlea (fluid)
Slide9THE MIDDLE EAR
The middle ear compensates for the impedance mismatch between the air in the ear canal and the fluid in the cochlea …
… by amplifying behaviorally relevant frequencies …Middle ear transfer function… except for a protective reflex that attenuates very loud sounds
Slide10THE EAR
OVERVIEW
Slide11THE COCHLEA
FREQUENCY ANALYSIS
The cochlea decomposes sound into its constituent frequencies
Slide12BM movement
Sound levelw/o OHCsThreshold for AN activation
Outer hair cells (OHCs) amplify weak sounds
Sound level
AN activity
w/ OHCs
THE COCHLEAAMPLIFICATION AND COMPRESSIONOuter hair cells are critical for sensitivity to weak sounds
Slide13THE COCHLEA
AUDITORY NERVE FIBER ACTIVITY
Temporal profile
Receptive fields
Kiang,
Acta
Oto-
laryng
, 1968
Phase-locking at low frequencies
Freq
= Joris et al, J Neurophysiol, 1994
Slide14The dynamic range of individual AN fibers is not sufficient …
Low thresholdHigh threshold
Sound level
AN activity
… but the collective dynamic range of the population is
THE COCHLEA
DYNAMIC RANGE FRACTIONATION
Slide15Signal from ear to brain
TimeAN activity
Time
AN activity
Signal from ear to brain
Incoming sound
Time
Sound level
Sound with increasing level in quiet environment
Time
Sound level
Incoming sound
Sound in noisy environment
Sound level
AN activity
Sound level
AN activity
THE COCHLEA
DYNAMIC RANGE FRACTIONATION
Slide16The cochlea is *not* just a frequency analyzer
DistortionCochlear position (mm)AN activity13.410.216.619.8
Signal from ear to brainSuppression
Amplification
Power
200040001000500Incoming soundFrequency (Hz)
THE COCHLEA
NONLINEAR SPECTRAL PROCESSING
Slide17Power spectrum of AN activity
PowerFiber BF ≈ F20.5
1.72.5Frequency (kHz)The vowel /e/PowerF1F2F3
Power spectrum of incoming sound
Young
, 2012Population representation“Synchrony capture” of vowel formantsTHE COCHLEANONLINEAR SPECTRAL PROCESSING
Slide18Outer and middle ear:Linear filtering
Cochlea:Frequency analysisAmplification/compressionDynamic range fractionationNonlinear transformation Phase-lockingThe signal from ear to brain is *not* just a spectrogram!AN activity
Time (s)Frequency (kHz)
0
0.5
11.5
22.542
0 A hu ge ta pes try hung in her hall wayThe ear encodes incoming sounds for transmission to the brain
THE COCHLEASUMMARY
Slide19Frequency
IntensityReceptive fieldTemporal profile
TimeActivity
sound
Low-D basis
Encoding
Ear
Feature extraction
Brainstem
Nonlinear recombination
MidbrainHigher cortexPerception
Selective amplification
Thalamocortical loop
High-D basis
High SNR basis
Classification
help
hello
yellow
mellow
Melo
Jello
fellow
THE MAMALIAN AUDITORY PATHWAY
OVERVIEW
Slide20THE COCHLEAR NUCLEUS
OVERVIEWLocation: medulla
Primary function: feature extractionMajor inputs: auditory nerve (E)Major outputs: superior olivary complex (E, from VCN)inferior colliculus (E, from DCN)Of note:many different cell typesDCN has complex micro-circuit
Slide21THE COCHLEAR NUCLEUS
MAJOR CELL TYPES
From AN
Slide22Receptive fieldTemporal profile
sound
THE COCHLEAR NUCLEUSMAJOR CELL TYPESVentral cochlear nucleus
Receptive field
Temporal profile
Dorsal cochlear nucleusDifferent cell types have distinct response properties
Slide23The outer ear creates elevation-dependent spectral notches
THE COCHLEAR NUCLEUS
MONOAURAL SPATIAL PROCESSINGThe DCN contains cells that are sensitive to the notch …… as well as the position of the head (and pinna)
Slide24THE SUPERIOR OLIVARY COMPLEX
OVERVIEWLocation: medulla
Primary function: (binaural) feature extractionMajor inputs: cochlear nucleus (E) Major outputs: lateral lemniscus (E)inferior colliculus (E)Of note: many understudied sub-nucleicalyx of held
Slide25THE SUPERIOR OLIVARY COMPLEX
THE CALYX OF HELDHundreds of active zones allow for fast, reliable transmission even at high input rates
The MNTB provides a sign change on the way to the SOC
Slide26ITD
ILD
Interaural
time difference (ITD)
Useful for frequencies < 2 kHz
Interaural
level difference (ILD)
Useful for frequencies > 2 kHz
BINAURAL SPATIAL CUES
Differences between the two ears indicate the position of sounds in the horizontal plane
Slide27ITDs are processed in the MSO …
THE SUPERIOR OLIVARY COMPLEX
BINAURAL SPATIAL PROCESSING… and represented by activity balance across hemispheres“labelled line” codebirds, lizards
“two-channel” codemammals
Slide28ILDs are processed in the LSO …
THE SUPERIOR OLIVARY COMPLEXBINAURAL SPATIAL PROCESSING
… and also represented by activity balance across hemispheres
Temporal multiplexing of acoustic and spatial information
Slide29THE LATERAL LEMNISCUS
OVERVIEW
Location: pons/midbrain borderPrimary function: sign change?Major inputs: cochlear nucleus (E) superior olivary complex (E)Major outputs: inferior colliculus (I)Of note: very understudied
Slide30Frequency
IntensityReceptive fieldTemporal profile
TimeActivity
sound
Low-D basis
Encoding
Ear
Feature extraction
Brainstem
Nonlinear recombination
MidbrainHigher cortexPerception
Selective amplification
Thalamocortical loop
High-D basis
High SNR basis
Classification
help
hello
yellow
mellow
Melo
Jello
fellow
THE MAMALIAN AUDITORY PATHWAY
OVERVIEW
Slide31THE INFERIOR COLLICULUS
OVERVIEWLocation: midbrain
Primary function: integration of brainstem inputs for relay to thalamocortical loopMajor inputs: cochlear nucleus (E) superior olivary complex (E)lateral lemniscus (I)Major outputs: medial geniculate body (E,I)Of note:no structure-function relationshipsno known micro-circuit
Slide32THE INFERIOR COLLICULUS
NONLINEAR RECOMBINATION OF BRAINSTEM INPUTS
0.5
0.5
59.5
0.5
58.7
0
0
0
0
0.1
0
0.2
0.1
0
0.2
0.2
0
0.5
0
0.5
16.0
0
0.5
19.8
0
0.01
0
0.2
4.7
0
0.2
10.6
0
0.2
0
0.2
2.0
0
0.2
5.5
0
0.5
0
100
Time (
ms
)
p (spike)
Responses to tones
10
80
250
4
Frequency (kHz)
Intensity (dB SPL)
1
0
0.5
0
0.5
Responses to speech
0
1
Time (s)
0
1
Time (s)
51.0
55.9
From
left
side
From right side
Receptive field
Temporal profile
Slide33THE INFERIOR COLLICULUS
HIGH TEMPORAL PRECISION
The neural basis for scene analysis has high temporal precision
Slide34Cochlear nucleusSpectral/temporal features
Monoaural spatial cuesSuperior olivary complex:Binaural spatial cuesInferior colliculusNonlinear recombinationThe basis for scene analysis is *not* just a spectrogram!The brainstem extracts useful features and combines them to create a high-dimensional neural basis for scene analysis
THE AUDITORY BRAINSTEM AND MIDBRAIN
SUMMARY
cochlear nucleus
superior olivary complexinferior colliculuslateral lemniscustrapezoid bodyexcinh
Slide35Frequency
IntensityReceptive fieldTemporal profile
TimeActivity
sound
Low-D basis
Encoding
Ear
Feature extraction
Brainstem
Nonlinear recombination
MidbrainHigher cortexPerception
Selective amplification
Thalamocortical loop
High-D basis
High SNR basis
Classification
help
hello
yellow
mellow
Melo
Jello
fellow
THE MAMALIAN AUDITORY PATHWAY
OVERVIEW
Slide36THE MEDIAL GENICULATE BODY
OVERVIEWLocation: thalamus
Primary function: attentional modulation?Major inputs: inferior colliculus (E,I)primary auditory cortex (E)thalamic reticular nucleus (I)Major outputs: primary auditory cortex (E)Of note: no interneurons in rodents
Slide37THE PRIMARY AUDITORY CORTEX
OVERVIEWLocation: Temporal lobe
Primary functions: attentional modulationcontextual processingMajor inputs: medial geniculate body (E)auditory and non-auditory (E,I?)Major outputs: higher-level auditory cortex (E)thalamus (E)Of note: ≥ 2 primary areas (A1, AAF, ?)
Slide38Rhythmic target
signal in noiseTimeFrequency
Power
Spectral filtering
Temporal filtering
Enhanced representationSpectrotemporal filtering improves SNR
THE THALAMOCORTICAL LOOPATTENTIONAL MODULATION OF SPECTRAL AND TEMPORAL SELECTIVITY
Slide39David et al, 2012
Sound discrimination task
THE THALAMOCORTICAL LOOPATTENTIONAL MODULATION OF SPECTRAL SELECTIVITYSpectrotemporal receptive fields in ferret A1Example neuron
Population average
Slide40Time (
ms re. sound onset)LFP amplitude (a.u.)Tone freq. = BFTone freq. = non-BF
Local field potential in monkey A1O’Connell et al., Neuron, 2011Single rhythmic tone stream - passiveFreq. = BFFreq. = non-BF
OR
~ 1 s
Lakatos et al., Neuron, 2013Dual rhythmic tone streams - active
Freq. = BFFreq. = non-BFAND~ 1 s
THE THALAMOCORTICAL LOOPATTENTIONAL MODULATION OF TEMPORAL SELECTIVITYTime (ms re. sound onset)Attend BF streamAttend non-BF streamLocal field potential in monkey A1
Slide41Spectral filteringAttention-dependent reweighting of inputs
Temporal filteringAttention-dependent entrainment of intrinsic fluctuations in excitabilityHigher-level cortex receives a modified neural basis that is suited to the current taskThe thalamocortical loop modifies the neural basis for scene analysis based on task-specific needs
THE THALAMOCORTICAL LOOPSUMMARY
Slide42Frequency
IntensityReceptive fieldTemporal profile
TimeActivity
sound
Low-D basis
Encoding
Ear
Feature extraction
Brainstem
Nonlinear recombination
MidbrainHigher cortexPerception
Selective amplification
Thalamocortical loop
High-D basis
High SNR basis
Classification
help
hello
yellow
mellow
Melo
Jello
fellow
THE MAMALIAN AUDITORY PATHWAY
OVERVIEW
Slide43HIGHER CORTEX
OVERVIEWThe auditory cortex plays an important yet ambiguous role in hearing. When the auditory information passes into the cortex, the specifics of what exactly takes place are unclear.
- WikipediaBizley and Cohen, 2013
Slide44HIGHER CORTEX
OVERVIEW
“Where”“What”Bizley and Cohen, 2013
Slide45+
=Reconstruction from cortical field potentialsSelective attention enhances the SNR of the neural basis for scene analysis
HIGHER CORTEX
NEURAL CORRELATES OF SELECTIVE ATTENTION
Mesgarani
and Chang, 2012
Slide46HIGHER CORTEX
CATEGORICAL RESPONSESNeurons represent sound class rather than acoustics in monkey STG
Time (s)Neural activityExample neuron
Morph (%)0
50
100Auditory task
Slide47Frequency
IntensityReceptive fieldTemporal profile
TimeActivity
sound
Low-D basis
Encoding
Cochlea
Feature extraction
Brainstem
Nonlinear recombination
MidbrainHigher cortexPerception
Selective amplification
Thalamocortical loop
High-D basis
High SNR basis
Classification
help
hello
yellow
mellow
Melo
Jello
fellow
THE MAMALIAN AUDITORY PATHWAY
OVERVIEW
Slide48HEARING LOSS
Death, taxes … and hearing loss
Slide49HEARING LOSS
Hearing loss makes it difficult to understand speech
Communication problems, social isolation, and more …
Associated costs in the US expected to exceed $50 billion annually by 2030 (Stucky, 2010)Hearing loss linked to increased:
Cognitive decline (Lin et al., 2013)
Dementia (Lin et al., 2011)Mortality (Contrera et al., 2015)
Slide50HEARING LOSS
High Low
Loud (74 dB SPL)
w/ aid
Background noise level
High Low% correct100500
w/o aidQuiet (52 dB SPL)Larson et al., 2000Hearing aids help, but not in noisy environments“I can hear you, but I can’t understand you”Two studies of speech recognition performance
Slide51BM movement
Sound levelw/o OHCsThreshold for AN activation
Outer hair cells (OHCs) amplify weak sounds
Sound level
AN activity
w/ OHCs
THE COCHLEAAMPLIFICATION AND COMPRESSIONOuter hair cells dysfunction decreases sensitivity
Slide52HIDDEN HEARING LOSS
Hearing loss also arises from changes to the changes to the AN itself
Cochlear synaptopathyProblem: high-threshold AN fibers are particularly vulnerableAN fibersIHC areaMild
SevereSynaptopathy in older people
From
Viana et al., 2015
Slide53Distortion
Cochlear position (mm)AN activity13.410.216.619.8
Signal from ear to brain
w/o OHCs
Suppression
Amplification
Power
2000
40001000500
Incoming soundFrequency (Hz)w/ OHCs
The cochlea is *
not*
just a frequency analyzer
THE COCHLEA
NONLINEAR SPECTRAL PROCESSING
Slide54Power spectrum of AN activity
PowerFiber BF ≈ F2Normal ear
PowerImpaired ear0.51.72.5
Frequency (kHz)The vowel /e/
Power
F1F2F3Power spectrum of incoming sound“Synchrony capture” of vowel formantsTHE COCHLEANONLINEAR SPECTRAL PROCESSING
Slide55Hearing loss is a profound distortion of the signal from ear to brain
Normal
ImpairedPopulation representation of /e/
THE COCHLEA
NONLINEAR SPECTRAL PROCESSING
Slide56HIDDEN HEARING LOSS
Hearing loss also arises from changes to the AN itself
Cochlear synaptopathyProblem: high-threshold AN fibers are particularly vulnerableAN fibersIHC areaMild
SevereSynaptopathy in older people
From
Viana et al., 2015
Slide57Sound level
AN activityHidden hearing lossLow thresholdHigh threshold
Low threshold
High threshold
Sound level
AN activity
Normal
Hidden hearing loss decreases differential sensitivity at high levels
THE COCHLEA
DYNAMIC RANGE FRACTIONATION
Slide58Signal from ear to brain
TimeAN activity
Time
AN activity
Signal from ear to brain
Incoming sound
Time
Sound level
Sound with increasing level in quiet environment
Time
Sound level
Incoming sound
S
ound in noisy environment
Normal
HHL
Sound level
AN activity
Sound level
AN activity
THE COCHLEA
DYNAMIC RANGE FRACTIONATION
Hidden hearing loss impairs perception at high sound levels
Slide59Hearing loss is much more than just a sensitivity problem… it is a profound distortion of the signal from ear to brain
… that hearing aids fail to correct
HEARING LOSS
Slide60Nicholas A.
LesicaEar InstituteUniversity College Londonwww.lesicalab.com
WHY DO HEARING AIDS FAIL TO RESTORE NORMAL AUDITORY PERCEPTION?
Slide6180
Level6040200140 dB SPL100120Busy restaurantOffice
LibraryHearing ThresholdJet engineConstruction siteRock concertSound
Slide62David et al, PNAS, 2012
Two different sound detection tasks
THE THALAMOCORTICAL LOOPATTENTIONAL MODULATION OF SPECTRAL SELECTIVITY
Slide63THE THALAMOCORTICAL LOOP
ATTENTIONAL MODULATION OF SPECTRAL SELECTIVITY
Spectrotemporal
receptive fields in ferret A1
Example neurons
Population averagesDavid et al, PNAS, 2012Approach taskAvoidance task
Slide64Time
Frequency
TEMPORAL COHERENCE
FOR AUDITORY SCENE ANALYSIS
Time
Frequency (KHz)
0
8
Time
Frequency (KHz)
0
8
Slide65Time
Frequency (KHz)08
TimeFrequency (KHz)0
8
Time
… in which temporal coherence indicates a common source
Frequency
The brainstem projects the signal from the ear into a high-dimensional space …
TEMPORAL COHERENCE
FOR AUDITORY SCENE ANALYSIS
Slide66