Diapositiva 1

1
HOW DO COCHLEAR IMPLANT PATIENTS PERCIVE THE
SOUND?SYNTHESIZING AUDIO SIGNALS ACCORDING TO
CODING STRATEGY AND FEATURES OF THE PATIENTS M.
Bastarrica (1,2), Angel de la Torre (2) ,Manuel
Sainz (3,4)(1) Dpt. Applied Technology, MED-EL
España (Spain) (2) Dpto. Electrónica y Tecn.
Comp. Universidad de Granada (Spain) (3)
Servicio ORL, Hospital Universitario S. Cecilio,
Granada (Spain)(4) Dpto. Cirugía y sus
Especialidades, Universidad de Granada (Spain)

• 1.- Introduction
• Cochlear implants provide hearing perception to
  patients affected by severe and profound
  hearing-loss. However, knowing how the patients
  perceive the sound is extremely difficult. This
  aspect becomes important for all the specialists
  involved in treatment and rehabilitation of
  cochlear implant patients
• For speech therapists, a precise knowledge about
  the hearing quality provided by the cochlear
  implant is important in order to plan the
  rehabilitation. It could also be useful in order
  to set adequate expectations.
• Understanding how hearing perception is
  influenced by insertion depth, interaction among
  different channels or duration of the deafness
  would be useful for surgeons, as it would ease
  the selection of the most appropriate device for
  each case.
• In order to set adequate parameters in the
  fitting map, it is important to know the
  relationship between fitting parameters and
  hearing quality. By understanding the effect of
  the modification of fitting parameters over
  quality of sound, the clinical engineer is able
  to optimize the perception provided by the
  cochlear implant.
• The aim of this work is to study how implanted
  patients perceive the sound. In order to do it,
  we simulate the mechanisms involved in the coding
  strategy and the stimulation of the cochlear
  nerve. According to this simulation, from an
  input audio signal we synthesize another audio
  signal which represent how it would be perceived
  by an implanted patient.

• 2.1.- Modeling the processor and the cochlear
  implant
• The first block allows to incorporate into the
  model specific information about the model of
  cochlear implant, the coding strategy, the number
  of active channels and the programming map.
• Most cochlear implant
• Divide the input signal into frequency bands by
  means of a filter bank
• The envelope is then estimated for each channel
• Stimulation levels are adapted to the specific
  dynamic range of each electrode of the implant,
  according to the programming map
• Finally, the pulses are generated for each
  channel according to a coding strategy and at a
  given stimulation range.
• There are several aspects to be considered in the
  model
• How many channels are there? (Number of inserted
  electrodes)
• How is the filter bank designed? (Frequency
  scaling, IIR, FIR or FFT filters, etc.)
• Method for envelope detection (rectifierLow-Pass
  or Hilbert transform)
• Map law mapping audio dynamic range into
  electrical dynamic range
• Differences between ideal fitting map and
  programmed fitting map
• Coding strategy (F0F2, F0F1F2, M-peak, N-of-M,
  CIS,...)
• Stimulation rate, etc.

2.3.- Synthesis of audio signal The scheme of
synthesis from the stimulation pattern associated
to different channels is this one
2.2.- Modeling the electrode neural ends
interface The second part of the analysis block
allows to incorporate information about the
insertion depth, the interaction among channels,
the tonotopic spectral resolution the index of
surviving neural ends, the intensity resolution
of the neural ends, etc. One key aspects in this
block is the distribution of the electric field
along the cochlea. The inter-channel interaction
can be modeled assuming that the electrode Nth
stimulates neural ends close to Nth electrode,
but also the neural ends corresponding to
electrodes (N1), (N-1), (N2), (N-2), etc. So,
there is a transfer of energy to each electrode
to the adjacent ones.

• 2.- Development of the analysis-synthesis model
• In order to simulate how cochlear implant
  patients perceive the sound, we synthesize audio
  signals taking into account different aspects
  involved in hearing perception with cochlear
  implants.
• The model of perception consists of two main
  blocks
• ANALYSIS This block process the audio signal
  from acquisition by microphone to pattern of
  activity in the cochlear nerve. It can be
  decomposed into two elements
• A technical block, involving the coding
  strategy, the cochlear implant processor and the
  electrode array of the cochlear implant.
• Electrode-neural ends interface block, involving
  interaction between the cochlear implant and the
  cochlear nerve.
• SYNTHESIS This block built an audio signal from
  the pattern of activity in the cochlear nerve.

In the simplest case, the filter bank should be
the same that was used for analysis. However, a
different frequency could be used to model the
insertion depth of each electrode (by using the
frequency associated to the allocation of the
electrode according to the tonotopic theory). As
excitation, a white noise could be used, modeling
a situation in which the neural activity is not
synchronized with the stimulation. A signal
consisting on a series of impulses can also be
used. This would model a situation in which the
neural activity is synchronized with the
stimulation. In that case, the impulses must be
computed from the stimulation pattern.

• 3.- Implementation and use of the model
• The described model have been implemented in C
  language.
• A software have been developed in order to
  synthesize audio signals according to the model
  parameters. The software can be download from the
  web http//www.ugr.es/atv
• It consists on a .exe file which can be run in
  a MS-DOS command window in a Windows-based
  system.
• Arguments are entered throughout the command line
  according to the following syntax (see more
  details, instructions and demos in the web)
• simulation_ci ltfile.wavgt ltrategt ltFmingt ltFmaxgt
  ltlength-CIgt ltN-chan-CIgt ltN-inserted-chgt ltN-on
  (N-of-M)gt ltinterac-decaygt ltHilbert(0) RLP(1)gt
  ltsynchron good(0) poor(1)gt
• ltrategt is the stimulation rate
• ltFmingt and ltFmaxgt define the frequency range
• ltlength-CIgt is the length of the electrode array
  in mm
• ltN-chan-CIgt and ltN-inserted-chgt are the numbers
  of channels and the number of electrodes inserted
  into the cochlea, respectively
• ltN-on (N-of-M)gt is the number of channels
  activated in each stimulation cycle for N-of-M
  based strategies
• ltinterac-decaygt is the decay constant (in mm)
  describing the inter-channel interaction
• ltHilbert RLPgt selects FIR filters with
  Hilbert transform envelope detection or IIR
  filters with rectifierlow-pass-filter envelope
  detection
• ltsynchronizationgt select synthesis in the case
  of good-poor synchronization of the activity in
  the cochlear nerve

The aim of this model is to allow the
incorporation of different aspects influencing
the hearing perception, including details related
to the processor, the cochlear implant,
the coding strategy, the fitting map, the
insertion depth, the inter-channel interaction,
the state of the neural ends, the interaction
electrodes / neural ends, etc. This way,
normal hearing subjects could hear the effect
of modifying each parameter (technical or
physiological) influencing the perception with
the cochlear implant.

• 4.- Conclusions and future work
• The simulation method is an useful tool for a
  better understanding of the influence of the
  different parameters (technical of physiological)
  involved in the perception of the sound by
  cochlear implant patients
• The software is currently under development.
  Only a part of the effects involved in the
  perception with cochlear implant have been
  included. There are many factors to be included
  in future revisions.