MRC Cognition and Brain Sciences UnitImproving hearing through a cochlear implant
How a cochlear implant works
People with a mild or moderate hearing loss can benefit greatly from a hearing aid, which amplifies sounds so that they can be heard. However, in some people, the receptor cells in the inner ear are so badly damaged that this amplification does not help. When this happens the patient is severely or profoundly deaf, and hearing can be restored by a cochlear implant (shown below). Sound is first picked up by a microphone worn behind the ear, and information about the sound transmitted to an internal receiver/stimulator using a radio-frequency link. The receiver/stimulator then sends trains of electrical pulses to electrodes inserted into the inner ear. These electrodes stimulate the auditory nerve, thereby bypassing the damaged receptor cells. Each electrode codes a particular frequency region of the sound, thereby mimicking the way that the healthy ear breaks sounds down into separate frequency components.
Cochlear implants are amazingly successful, and have restored hearing to more than 200,000 deaf patients worldwide. However, even the most successful patients have difficulty in perceiving the pitch of sounds, leading to great difficulty both in enjoying music and to understanding speech when more than one person is talking at once. This can be illustrated by listening to the following demonstrations, which simulate the way in which sound is processed by a cochlear implant.
- Listen to this sentence (listen). Although it sounds a little un-natural, it is possible to understand what is being said, especially when you have some idea of what is being said (listen)
- Now listen to this piece of music (listen). Even when you know the words (listen), you probably don’t enjoy it very much
- Listen to these two people talking at the same time (listen). It is possible to attend to either the man’s or the woman’s voice (listen), by focussing on either the lower or the higher pitch. However, because cochlear implants don’t encode pitch very well, patients find it hard to separate the competing voices (1)
Improving pitch perception and sound segregation by cochlear implant users
CBSU scientist Bob Carlyon and his team are investigating ways of improving pitch perception and sound separation by cochlear implant users. For example, we have developed a method for improving the range of pitches that implant users can hear (1), and, with colleagues at the University of Leuven, are investigating ways of more selectively stimulating different auditory nerve fibres. Bob’s recent worked, funded by the charity Action on Hearing Loss, aims to identify electrodes that do not effectively convey acoustic information, and of using this information to re-program the implants of patients who are having difficulty understanding speech. These projects are all carried out in close collaboration with clinical colleagues at Addenbrookes hospital (see CHIRP) and with the help of the major cochlear implant manufacturers.
Improving speech perception by Auditory Brainstem Implants
When the auditory nerve is damaged – for example as a result of a tumour – cochlear implants cannot restore hearing. Instead, it is possible to stimulate the next stage of the auditory pathway with an Auditory Brainstem Implant (“ABI”). Unfortunately, speech perception through an ABI is usually much worse than with a cochlear implant. One reason is that, unlike with a cochlear implant, it is not possible to know in advance which electrodes will produce a high pitch and which will produce a low pitch. We have developed a statistically based method that allows the audiologist to determine the correct ordering in as short a time as possible (2). This method is in use in clinics both in the U.K. and the U.S.A. We are currently performing further experiments into the limitations of ABIs, with the aim of producing recommendations for how they are programmed.
Alleviating tinnitus in cochlear implant users
A common treatment for tinnitus is to listen to recordings of natural environmental sounds, such as wind blowing through the trees or a trickling stream. Unfortunately, when we played these sounds to cochlear implant listeners, they did not sound natural, and the listeners often found it hard even to discriminate between the sounds. We then found that this was because cochlear implants do not effectively convey the (sometimes fast) modulations in sound amplitude that characterise different environmental sounds. Our colleague Dr Rich Turner, in the Cambridge University Engineering department, has analysed the statistics of environmental sounds, and has developed a computer model that can use these statistics to generate novel examples of, for example, wind or of running water. We have used the model to generate a wide range of new sounds that have some of the statistical characteristics of environmental sounds, but which are easily discriminable from each other. So far we have obtained a good idea of what differences are and are not audible to implant users. Our next step is to develop a software tool whereby implant users can adjust the different statistical properties of sound in order to identify the stimulus most likely to alleviate their tinnitus. This research is being carried out by CBSU Ph.D. student Phil Gomersall, jointly supervised by Bob Carlyon and Dave Baguley (Head of Audiology, Addenbrookes hospital).
(1) Extending the Limits of Place and Temporal Pitch Perception in Cochlear Implant Users
OLIVIER MACHEREY, JOHN M. DEEKS, AND ROBERT P. CARLYON
J. Assoc. Res. Otorhinolaryngol. 12, 233-251
A series of experiments investigated the effects of asymmetric current waveforms on the perception of place and temporal pitch cues. The asymmetric waveforms were trains of pseudomonophasic (PS) pulses consisting of a short, high-amplitude phase followed by a longer (and lower amplitude) opposite-polarity phase. When such pulses were presented in a narrow bipolar (“BP+1”) mode and with the first phase anodic relative to the most apical electrode (so-called PSA pulses), pitch was lower than when the first phase was anodic re the more basal electrode. For a pulse rate of 12 pulses per second (pps), pitch was also lower than with standard symmetric biphasic pulses in either monopolar or bipolar mode. This suggests that PSA pulses can extend the range of place-pitch percepts available to cochlear implant listeners by focusing the spread of excitation in a more apical region than common stimulation techniques. Temporal pitch was studied by requiring subjects to pitchrank single-channel pulse trains with rates ranging from 105 to 1,156 pps; this task was repeated at several intra-cochlear stimulation sites and using both symmetric and pseudomonophasic pulses. For PSA pulses presented to apical electrodes, the upper limit of temporal pitch was significantly higher than that for all the other conditions, averaging 713 pps. Measures of discriminability obtained using the method of constant stimuli indicated that this pitch percept was probably weak. However, a multidimensional scaling study showed that the percept associated with a rate change, even at high rates, was orthogonal to that of a place change and therefore reflected a genuine change in the temporal pattern of neural activity.
(2) C.J. Long, I. Nimmo-Smith, D.M. Baguley, M. O’Driscoll , R. Ramsden , S.R. Otto, P.R. Axon, and R.P. Carlyon (2005). “Optimising the clinical fit of auditory brainstem implants”. Ear and Hearing. 26, 251-262
Objective: To develop and implement a new audiological fitting procedure for auditory brain stem implants (ABIs), based on an efficient algorithm, and to compare it with two procedures presently used in clinical practice. Design: First, the different procedures were compared by using computer models and simulations with normal-hearing subjects (N 4). This allows for an analysis of the accuracy of the procedures in a way that is not possible when testing ABI users. The root-mean-square error between the order estimated by the procedure and the true order was calculated. In addition, ABI users (N 2) were tested with the new procedure to see if it could be successfully applied in clinic. The degree of variability of their results across runs and sessions was analyzed. Results: The tests of the normal-hearing subjects showed that our proposed procedure required significantly fewer trials (22 on average) than procedures presently used in clinic (with 76 and 234 trials on average for the two other procedures tested) to produce the same degree of accuracy. Computer modeling also demonstrated this advantage. Additional testing showed this advantage was maintained under a variety of conditions relevant to the clinic. The two patients tested were able to use this procedure with success, even though they were poor at discriminating the pitch of electrodes. The patients showed results consistent with having about 4 to 5 discriminable groups of electrodes with the 12 to 14 electrodes tested. Conclusions: The proposed procedure requires fewer trials to produce a clinically useful result and is well tolerated in the clinic. An additional advantage is that it allows testing to be broken down into several “blocks,” each containing a small number of trials. If the variability between blocks is small, information can be combined across blocks to increase the accuracy of the result. If the variability is large, perhaps between blocks on different days, this may reflect a significant change in the percepts generated by the implant, and signal to the clinician that a significant alteration in the fitting is required. We recommend its use in ABI user fitting and in cochlear implant fitting when pitch ranking is problematic.
