Opti-Fox
The Opti-Fox consortium attempts to contribute to the quality improvement of cochlear
implant fitting.
CI fitting is the process of programming the implant to the individual anatomy,
physiology and other patient-related parameters. To date this is a very tedious
and laborious job which is executed by highly expert audiologists, engineers, medical
doctors etc. With more than 200.000 CI users worldwide and an annual increase of
over 30.000, finding the skilled professionals and the time to perform the fitting,
becomes an ever-increasing problem and a real bottle neck to the further implementation.
On top of this, CI fitting as a process is not yet at the stage of being well controlled,
standardized or mature as professional discipline. Huge variation exists between
the different approaches of different experts. No universal guidelines exist, no
formal training, no established “Good Clinical Practice”.
The Eargroup has been addressing these issues since many years. Process optimization
requires valid tests to monitor outcome and systematic procedures to standardize
and drive the process.
- Since audiometry and speech audiometry were judged insufficient to monitor the coding
of the different components of sound, A§E has developed. This is a psychoacoustical
test suite that is language independent and feasible in the clinical practice. Different
test modules have been and are still being developed to assess the coding of intensity,
spectral and temporal content of sound at the level of detection, discrimination
and identification. See A§E for more details
and instructional videos.
- FOX (Fitting to Outcome eXpert) is a software tool under development to systematize
the fitting procedure. In its present form, it is a rudimentary intelligent agent
that runs deterministic logic (the “Eargroup Advice”) to read and interpret the
cochlear implant program and the measured psycho-acoustic outcome obtained with
this program. Based on this analysis, it executes the advice and proposed changes
to the CI program. The audiologist can accept these changes and program them to
the CI speech processor. FOX is currently being investigated and validated in clinical
trials in several CI centres in Europe and India.
Opti-Fox focuses on both pillars of the Eargroup approach.
- Outcome measuring:
- If more emphasis is going to be put on psycho-acoustic outcome measurements, it
is of paramount importance to create state of the art and well calibrated test conditions.
Good criteria and working principles have been defined by several international
standards. At this moment they require expensive and space-taking sound-treated
rooms and equipment. Opti-Fox will try to create other test conditions that assure
the same acoustic test quality while reducing the cost and required space substantially.
Input from engineers and acousticians is crucial for this enterprise.
- Speech audiometry remains a cornerstone in the monitoring of CI performance. This
test, however, is time-consuming, language-specific and highly dependent on the
linguistic skills of the test person. These are serious drawbacks for the universal
usability of speech audiometry for the purpose of process-optimization. An attempt
will be made to reduce these drawbacks by making the test less language-specific,
more adapted to the individual linguistic skills and less dependent on professional
time. Input from linguists and speech technologists is crucial for this enterprise.
- Process automation
- At this stage, FOX is a rudimentary piece or artificial intelligence (AI). Opti-Fox
will attempt to improve the level of AI by modeling the functional relations between
the many variables involved and by adopting one or more self-learning strategies
to improve the accuracy of the advice given. Input from computational technology
and AI specialists is crucial for this enterprise.
Update Sept 2011
The project activities have been divided into 4 Work Packages:
- WP 1: Language-independent Speech-(in-Noise) testing
- The first task of this WP consisted in a detailed state-of-the-art with respect
to speech audiometry in the native languages of the SME- and RTD-partners involved
in the OPTI-FOX consortium. Research consisted in a presentation and critical discussion
of existing tests for Dutch- and German-speaking testees in view of the criteria
and norms of the word lists used in these tests.
- Secondly, recordings were made of speech samples of typical and atypical (e.g. hearing
impaired) listeners in view of the development of a tester-independent speech audiometric
test battery. In order to be able to develop a language-independent template for
this test battery a metric was established to determine the representativeness of
a speech or text sample of a given language, based on the linguistic features of
that language. As the metric will be used to build lists of words that serve as
acoustical prompts in speech audiometry, priority was given to a distance measure
evaluating the use of graphemes to represent the sound system of a given language.
- Finally, this WP was also concerned with the development of a software engine to
segment text into word lists, apply filters on charachter count and present these
words visually for use in WP2.
- WP 2: Objective automatic speech error analysis
- The aim of this WP is to introduce automated speech analysis in speech audiometry
testing. Firstly, a number of concerns regarding accuracy and robustness of such
an analyses that might compromise its validity in the present medical context, have
been addressed. The focus has been mainly on inter-person variability analysis in
order to design an assessment method that is able to detect differences between
two acoustic realizations of the same word spoken by a single speaker. Available
techniques were explored in order to provide a 'dissimilarity measure' for the difference
between them.
- Secondly, this WP was also concerned with the development of a software engine to
record the words from WP1 as wave files after visual and acoustic prompting and
to analyze these wave-files by means of common ASR technology. At the conclusion
of this reporting period, some 180.000 wave files from more than 200 different speakers
(Flemish, Dutch, German) have been recorded and analyzed.
- WP 3: Optimization of the automated fitting process for
cochlear implant speech processors
- Within this WP new powerful methods, algorithms and software tools for tuning cochlear
implants are being built that are intended to overcome limitations of manual trial-and-error.
The first stage of the project consisted in problem understanding, data collection
process, data formats and an exploratory analysis of available data.
- The second stage consisted of data modeling and the design of the application. Relations
were captured between electrical input (map settings) and psycho-acoustic output
(outcomes of tests) variables. The key question to be answered with help of data
modeling was which cochlear implant parameters should be changed, and by how much,
in order to obtain a desired change in the results of a specific test. Currently,
several approaches to develop an optimal tuning strategy are being explored, ranging
from general optimization methods to the most recent methods that were developed
in the field of active learning.
- WP4: Test box for cochlear implant testing
- The goal of this WP is to develop a test box with ideal conditions for the measuring
of psycho-acoustic outcomes with cochlear implants. The box should be able to (i)
exclude external noise (acoustic insulation) as defined by ISO standards, and (ii)
to produce an exact acoustic replicate of an electric input signal (a simulation
of ANSI Type 1 free field testing). The requirement with respect to the acoustic
insulation of the test box focused on efficiency within the set limits of weight,
size and cost price. In a first stage, an inventory was made of available norms,
specific requirements and possible technology to be used. In a second stage a first
prototype was build.
Expected final results
We expect to develop 3 products which together will drastically change the technical
fitting of cochlear implants:
OTOspeech aims to be a psycho-acoustical test to assess the auditory performance
of cochlear implant users in a universal and automated way; universal in that it
will be usable in a huge number of different languages and that it will be independent
of dialect of lexical knowledge; automated in that the scoring of the speech intelligibility
will be done in an automated way; this will allow the apllication to run as a self-test.
FOX will be optimized to become an AI (artificial intelligence) application which
will analyze the electrical parameters of a cochlear implant together with the psychoacoustical
test results obtained with is and which will produce recommendations for changing
the electrical parameters such that the outcome will improve to target; this application
will be optimized by modelling the many input variables as a function of the many
output variables and by using AI technology (the currently selected candidates are
surrogate modelling, nearest neighbour analysis and Bayesian technology). We will
undertake to develop the final application in such a way that it will be self-learning
by continuous analysis of the growing data set.
The FoxBox intends to be a portable desktop module with integrated amplifier, loudspeaker
and soundcard, which will allow the psycho-acoustic testing and fitting of CI-users
in standardized and calibrated conditions without the need of sound-treated test
rooms with highly specialized and expensive audiological equipment.