IM this article to a friend!

March 3, 2004

Cochlea logic

From: Radio Netherlands - Netherlands - Mar 3, 2004

by Laura Durnford of our Science Unit, 3 March 2004

For people who become deaf after years of normal hearing, a cochlear implant can be a mixed blessing; it does restore the ability to hear, but the 'sound' can be very distorted and difficult to adjust to. Recent research shows the brain's ability to adapt, and should help to improve not only the performance of the implants, but also the quality of life for the patients.

A snail-shell shaped organ in the inner ear is the part that does the business of converting the vibrations of sound waves into electrical signals, with different frequencies being processed at different points along the length of the coil.

The signals are then conducted along the auditory nerve and interpreted by the brain. When this organ, the cochlea, goes wrong, the brain can no longer receive auditory information and deafness is the result.

For around 25 years, it has been possible to re-establish the connection between the brain and the audible world, using an implanted electronic device to artificially stimulate the auditory nerve and restore the ability to hear.

Dr Robert Shannon of the House Ear Institute in Los Angeles, USA, comments:

"The average [cochlear implanted] person is able to understand about 90 percent of words and sentences under good conditions – clear speech, face to face communication, no noise in the background – and most of these people can talk on the phone without much difficulty. But we're still struggling to find the best way to tune the implant to customise it for an individual person."

Quality limitations

Cochlear implants comprise of two main parts, a receiving device worn externally either behind the ear like a normal hearing aid or on a strap like a walkman, and the implant itself. This is a wire-like object about 30 millimetres long and one millimetre in diameter with up to 20 electrodes along its length. The external receiver processes sound into electrical signals, which are then fed into the auditory nerve by the electrodes.

There are technical limitations to the quality of the signal that a cochlear implant can send. The first problem is that there is only a low number of electrodes to channel the signals, compared with a healthy, hearing ear. This results in so-called 'spectral smearing', where details get lost in a coarser representation of the original, a bit like a pixelated picture (see sample).

Frequency mismatch

The second major problem is 'frequency mismatch'; each electrode conveys only a particular band of frequencies and if this doesn't match up with the area of the cochlea where that frequency was previously processed, the signal sounds distorted. A visual analogy would be to twist a picture (see example) and this can happen if the device is not inserted far enough into the cochlea during surgery.

So what can scientists do about these problems? Spectral smearing cannot be helped, because of the physical limit on the number of electrodes that can be included in the device. But frequency mismatch is another story, as Professor Mario Svirsky from the Indiana University School of Medicine explains:

"Once the electrodes are inside the cochlea... there will be an amount of frequency mismatch that may be small or may be significant. That's something we have no control over once the surgery is over. However, we do have control over the speech processing done by the external equipment".

So by 'tuning' the processor in different ways, researchers hope to help cochlear implant recipients to cope better with their new hearing.

The cochlea's geography Dr Stuart Rosen is Professor of Speech and Hearing Science at University College London, in the UK. He has been investigating what may be the best approach in dealing with frequency mismatch. He says that high frequency sounds, such as a hissy 's', are generally processed in the base of the cochlea, while low frequency sounds are dealt with towards the apex of the spiralled organ, so important low frequency sounds that give vital information about the meaning of words are lost if the implant fails to penetrate to the tip of the cochlea. "There are really two ways you can deal with this," he says:

"You can say 'you must match the frequency band to the place it normally goes in the cochlea'. You just have to lose the information that's contained in the low frequency sounds of speech. Or what you can say instead is, 'I want to pick the frequency regions I know are most important for speech and put them on whatever electrode array I have, however deep it goes'." This means that "speech now sounds to be much higher in frequency than it normally is" and this 'frequency shifted' signal can also be difficult to understand.

Teaching the brain

By simulating the effects of frequency shifted speech and playing it to hearing people, Professor Rosen has come to the conclusion that the second option would be preferable, because experience teaches the brain to find the necessary lower-frequency information within the distorted signal. But the process of adaptation is a slow and difficult one. It can take about a year of hearing distorted gibberish before implant patients achieve a reliable rate of success in understanding it. So, would training help?

Professor Mario Svirsky has been finding out, again by testing how well hearing volunteers understood simulated cochlear implant signals. One group underwent the 'standard treatment', hearing sentences that were both spectrally smeared and frequency shifted (see and hear example) and being left simply to try to understand.

A second group received the 'experimental treatment', beginning with spectrally smeared sound only, so losing out on important low-frequency information, but gradually being introduced to the richer frequency shifted signals as well. "Once the frequency shifting was introduced the experimental group did worse," he recalls, "but not a lot worse, whereas for the standard group the decrement was tremendous. They went down from more than 70% correct to slightly over 20% correct. You can also see that it took several training sessions for the standard group to catch up."

Seeing what you hear

By using imaging techniques to watch their subjects' brains in action, Professor Svirsky and his colleagues also saw results consistent with their behavioural tests. At the start and end of the experiment they used functional magnetic resonance imaging techniques to assess the response of two known language-processing regions of the brain when distorted sounds were played forwards and backwards.

"Prior to training there's little or no difference between those two conditions, which makes sense because it sounds like gibberish whether you play it forwards or backwards," laughs Professor Svirsky, whereas after 15 hours of training, "we see that there are differences here when we play the signals forwards and backwards."

So by adjusting an implant's sound processor unit, it should be possible to let patients first become accustomed to the spectral smearing effect before adding in the richer, but potentially confusing, frequency shifted signals – so allowing them to get as much information about the original signal as possible. Professor Svirsky stresses that his findings suggest that such training "would not result in better long term results, but it would result in quicker adaptation to the unnatural signal." And he adds that by reducing the process to a month, rather than a year, "that would result in tremendous improvement in quality of life" for these patients.

Pictures of Abraham Lincoln courtesy of Prof. Svirsky


Prof. Svirsky's Department at Indiana University School of Medicine

Dr Robert V. Shannon's Homepage

Prof. Stuart Rosen's Homepage

More on cochlear implants

© 2004 Radio Netherlands