Cochlear Implants: My Perspective

By William F. House, D.D.S., M.D. (Edited by David House)

When a subject is highly controversial
one cannot hope to tell the truth.
One can only show how one came to hold
whatever opinion one does hold.
One can only give one's audience the chance of
drawing their own conclusions
as they observe the limitations,
the prejudices, the idiosyncrasies of the speaker.

~Virginia Woolf, A Room of One's Own

Chapter II, Controversy, or: How Do Implants Work?

EVEN WITH THE APPARENT broad acceptance of implants, controversies, remain. The most important of these, as I indicated in the introduction, is the apparent conflict between clinical observation and theory.

The Tonotopic Theory

The work of von Bekesy [22] in the thirties and forties firmly established the traveling wave theory and the concept of discrete areas of stimulation along the basilar membrane. That is, von Bekesy made it clear that the design of the cochlea "sorted" sounds along its length so that in response to a particular frequency, a specific area of the basilar membrane exhibited the greatest vibration, and correspondingly the hair cells (and their attached dendrites) present at that site were stimulated. This is referred to as tonotopic stimulation, as previously explained.

The key point is that it requires stimulation be site-specific, and limited in its spread. Therefore, the tonotopic theory assumes that the hair cell is the only site of stimulus for the dendrite. In this regard, I would note that because an VIII nerve fiber is a bipolar neuron, it has only one dendrite. (Its spiral ganglion cell body is in Rosenthal's canal in the modiolus.) [23] (See Figure 1)

In any case, on the basis of these assumptions, we reasoned that if discrete clusters of dendrites could be stimulated by placing multiple active and ground electrodes close together along the basilar membrane, the 'natural function' (that is the frequency and intensity discrimination of the cochlea) could be better duplicated. (See Figure 2)

So it was, as I mentioned previously, that in the late sixties, Jack Urban and I began working with five-electrode, hard-wired systems placed in the scala tympani of totally deaf volunteers. I wish to emphasize that our reasoning was entirely molded by the tonotopic theory: it guided our every step, in the beginning. Thus, in early work done with cochlear implants, as mentioned in the previous chapter, our idea was that in total deafness the hair cells were missing, but that the dendrites within the basilar membrane and their associated spiral ganglion cell bodies remained intact, and further that it was through the direct stimulus of the dendrites on the basilar membrane that the electric impulses provided by the implant were transduced into nerve impulses. This was why the first hard-wired implant was "multi-channel".

Many different electrical inputs were tried, but to our amazement putting the same signal into all the electrodes simultaneously seemed — as reported to us by the patients — to give the same or better sound perception as the multiple, limited-area inputs.

This of course was totally unexpected and very confusing. We could not explain it, and it appeared that our reasoning by analogy with what we knew about the cochlear processing of pure tones had somehow provided us with incorrect assumptions. In other words, the theory of tonotopic stimulation which we explored in this early work, as attractive as it was, and regardless that it fit well with what we knew about the natural function of the cochlea, did not, in practice, work.

It was for this reason that when Jack Urban and I developed the original, totally-implanted hardware, a single electrode was used: every test which we had devised demonstrated to us that this was what was needed. The first 10 patients implanted with this simple electrode were among those that went to Pittsburgh for the Bilger study.

It is ironic that between the tonotopic theory and this clinical work, which clearly demonstrated that our theory was wrong, the theory should still have so much apparent potency and wide currency.

Anatomic evidence

But there is now much more data that bears on the matter. The studies of Linthicum et al [24] have clearly shown that in 16 temporal bones that had been implanted with cochlear implants, some had very few, and most had no basilar membrane dendrites. (See Figure 3)

Again, if the tonotopic theory is valid, then stimulable dendrites would have to remain in the basilar membrane: but these dendrites are almost always missing. [25]

This study unequivocally demonstrates that the most probable site of electrical stimulation in the cochlea is the spiral ganglion cell bodies. As such, the clinical data gathered in early work done with implants which puzzled us so completely is now independently supported by these anatomical studies. That is, there were no local dendrites to stimulate, so attempts at local stimulation were useless, and not as well received by the patients as more general stimulation.

(We are of course dealing with pathology here, and thus von Bekesy's concepts of cochlear stimulation remain important, and offer some idea of how a pure tone reacts mechanically within the cochlea. However, we are as yet ignorant regarding the mechanisms by which complex, supra-threshold sounds such as speech are handled by the cochlea. As well, we have virtually no insight into how the derived nerve impulses are processed by the brain and thereby become perceived sounds. Fortunately, cochlear implants are providing some new ideas about mechanisms of cochlear function.)

Local stimulation requires small fields

Certainly, the information presented above offers compelling reasons for discarding the tonotopic theory. Beyond this however, further clinical investigations using different electrode placements have offered clear evidence that the tonotopic theory was invalid in the context of implants.

Consider that local stimulation requires closely spaced and paired active and ground electrodes, and as well, it requires that small electrical fields be generated between these electrode pairs. Small fields in turn must be generated by small electrical potentials between the paired electrodes. Note that any enlargement of the field of electrical stimulation, whether caused through the use of a common ground ("de-pairing" the electrodes) or through the use of higher potentials (creating a more intense and larger field) necessarily engulfs more structures, thus widening the area of stimulation from the presumed target: the dendrites on the basilar membrane. If one wishes to be religious in one's reliance on the tonotopic theory, one must likewise be rigorous in using small fields and paired electrodes placed near the proper spot on the basilar membrane: there is no reasonable alternative.

Yet larger fields are more efficient

While no one knows enough about the electrical characteristics of the cochlea to predict precisely what the current levels "should be", nor the exact pathways through which artificially induced currents flow within the cochlea, it is clear that, for a given pure tone, local stimulation — if possible — will require less current to attain the threshold of perception than the amount of current required to electrically saturate the cochlea with a single, larger field. In other words, this comparison offers its own benchmark.

What then has the clinical evidence shown? Is the theory borne out in practice? Are measured charges lower with closely paired electrodes, as we would expect, thinking tonotopically?

Our early clinical findings demonstrated that, when stimulating between two closely-spaced electrodes in the scala tympani, more current was required to reach threshold than was required to saturate the cochlea generally. When using a single active and a single ground electrode, placed in such a manner that the current was directed across the modiolus and the spiral ganglion cells, lower currents allowed us to reach threshold.

But of course, this evidence comes from our early results. Is there other evidence which indicates that larger, non-tonotopic fields offer better results?

Of course, there is. In fact this inverse relationship between local stimulation and current requirements appears to be a general finding, and is not restricted to a single electrode system design. In the April 1993 edition of the Cochlear Corporation's Audiologist's Handbook for the Nucleus 22 implant, on page 149, it states:

Typically, the lowest thresholds are obtained in CG [common ground] because of the wider current spread.

"CG" for this device obtains when one electrode is active, and all 21 others are being used as ground electrodes. Note that this "breaks" close pairing between electrodes, and creates a larger field. The chart which is found on the same page of the Handbook (their Figure 41) offers a graph of experimental data which shows that, using common ground in a saline solution (which was likely used due to its presumed electrical similarity to the electrolytic fluid of the scala tympani) the electric field is rather evenly and equally spread from the active to all other electrodes, much as water would drain equally through similarly-sized holes in the bottom of a can.

As the quote makes clear, it "typically" requires less current to achieve threshold when electrode pairing is broken, and the field of stimulation spreads. Indeed, setting the Nucleus system up to use a common ground is a popular practice. [26]

Recently I was told that Cochlear Corporation will in the future use a ground outside the cochlea to further cut their power requirements. Common experience is that batteries are wearing out too fast with closely paired stimuli, because current requirements are higher.

Stimulating the spiral ganglion cell bodies

Thus all available evidence points in the same direction: Larger fields are required, and smaller fields will not work.

For example, consider that closely paired electrodes require more current for a given effect. This again demonstrates that closely-paired electrodes must generate more intense fields in order to spread the stimulation: a logical contradiction between design and practice.

With closely paired electrodes, most of the current tends to flow in the most direct path between them. Thus, using such closely-paired electrodes, a more intense current must be passed between them, such that the "spill-over" is sufficient to generate a larger field. This is the sort of effect we might predict if we believed precisely the opposite of the tonotopic theory, to wit, that local stimulation, far from being required, was in fact quite useless, and that, when constrained by designs which direct the current to flow locally, we must "waste" enough current that the excess current will produce the desired result.

But if we are not stimulating local dendrites on the basilar membrane, what then are we stimulating?

Our belief is that the most probable site of stimulation is the spiral ganglion cell bodies within the modiolus. Those familiar with the anatomy of the cochlea will have difficulty choosing another potential site. The conclusion, once again, is that the most probable way — indeed one may say the only possible way — to cause the remaining neural structures of the damaged cochlea to fire is to blanket the spiral ganglion cells of the VIII nerve with an electric field.

The generation of this field within the cochlea is most efficiently accomplished (that is, accomplished with the least electrical voltage to the electrodes), when the ground electrode and the active electrode are placed such that the field generated will pass through the cochlea. (See Figure 4)

The Morse code rationale

We would expect, based on current neurophysiologic understanding, that any electric field, when strong enough to achieve threshold, would cause indiscriminate firing of all the cochlear neurons: what would prevent that? Further, the expectation would be that the neurons would then go into a refractory state, as previously described. So, conventional wisdom tells us, all the patient should hear something like an 400 Hz buzz whenever any current capable of stimulating the nerves, regardless of its other characteristics, is applied generally to the cochlea.

This theoretical model of neural activity in response to artificial stimulation makes so much sense, in fact, that it is difficult to understand that it is not correct, or why. So the mystery is deepened, and the challenge is widened. Not only do the facts which confront us require us to let go of the tonotopic theory, but as well they bring into question some of our basic neurophysiologic expectations.

It should be very clear that as yet we do not understand enough about how the normal auditory system processes pure tones, much less to figure out how implants work. [27] Rather, we must judge cochlear implants by their results through clinical observation, and not be prevented from trying different approaches because preconceived and unfounded notions say something cannot work. Otherwise, we are back to the days of Galileo, battling with the ghost of Aristotle about which iron ball will fall faster.

Finding wiggle room

Still there are those who are unwilling to completely abandon the theory of tonotopic stimulation within the cochlea. Rather than believing that implants operate by stimulating dendrites along the basilar membrane, however, their feeling is that what is required is to selectively stimulate the spiral ganglion cell bodies (within Rosenthal's canal in the modiolus) which apparently correspond to a certain frequency.

It should be remembered, however that the spiral ganglion cell bodies are closely packed, and are surrounded by blood vessels and fluid spaces. The active electrodes are surrounded by the electrolytic fluid of the scala tympani. This fluid seems ideally suited to conduct the electric fields generated by the cochlear implant widely and generally. Under these circumstances it would seem completely illogical to assert that a current is somehow finding its way to, and stimulating only certain discrete numbers of spiral ganglion cells in exclusion to any others. Further, given that the amount of current required to reach threshold in production implants is lower when using a common ground or a remote ground outside the cochlea, it seems by extension illogical to believe that the tonotopic theory can be salvaged by assuming that discrete sets of dendrites along the basilar membrane have somehow been replaced by discrete sets of spiral ganglion cell bodies within the modiolus, as the site of stimulation.

The geography of frequency

Independent studies further demonstrate that it is unnecessary to generate electric fields across either the basilar membrane dendrites or the spiral ganglion cell bodies from different sites in the scala tympani.

Consider that, as we currently understand it, the frequency sensitivity of the basilar membrane moves from higher frequencies near its beginning at the round window, to lower frequencies as one approaches its end at the apex. As such, we can — at least theoretically — roughly correlate distance along the basilar membrane with an associated site of stimulation for a given frequency.

So we can say that a 25 mm electrode (e.g. of the length used by the Nucleus 22-channel implant or the Ineraid implant) inserted into scala tympani from the round window will extend from 20,000 Hz at the round window up to about the 1500 Hz area. By contrast, a 6 mm single electrode (e.g. of the length used by the 3M/House and the AllHear implant) extends only up to about the 4000 Hz area. Indeed, because stimulation via multiple electrodes is intended to be local to frequency-specific areas of the basilar membrane, then the electrode must be as long, or nearly as long, as the basilar membrane itself. That is, tonotopically-correct electrodes must necessarily be long electrodes. In fact, given that there are no electrodes in any in-production implant which reach to areas on the basilar membrane of 1500 Hz and below, it can also be said that there is, today, no such thing as a tonotopically-correct electrode. We are all, perforce, practicing this particular sin of omission.

The psychoacoustic studies of Bilger of the first 12 single 6 mm electrode patients, found that these patients had normal frequency difference limens in two cases up to 2000 Hz, and in the remaining cases up to 500 Hz. [28] That is, sounds with frequencies below 2000 Hz are accurately discerned despite the fact that active electrode could not possibly "directly" (tonotopically) stimulate the 2000 to 250 Hz frequency-related area of the basilar membrane. Under these conditions, where electrodes cannot stimulate the "correct" spot on the basilar membrane, it would of course be impossible for either the basilar membrane dendrites, or the corresponding spiral ganglion cell bodies, to be locally, discretely, and exclusively stimulated. One must as a result either recognize that the tonotopic theory in all its potential variations is invalid, or reject the clinical data.

It is now commonly observed that patients with short single 6 mm electrodes, and those with multiple long electrodes in the scala tympani have equivalent pure tone thresholds from 250 to 8000 Hz, [29] thus reinforcing the understanding that there is no need to stimulate from "tonotopically correct" sites in the scala tympani.

In sum, we now know with certainty that long multi-electrode cochlear implant systems do not — indeed, the cited studies show that they most often cannot — do what they were designed to do.

A death in the family

In sum:

• Anatomic studies disallow the tonotopic theory: in most cases no local dendrites survive on the basilar membrane, making local stimulation impossible.
• If fields are localized by electrode design, practice demonstrates that relatively more current must be used to achieve a given effect. Logically speaking, this can only be the case if the field is forced to grow until it stimulates more structures, and thus stimulates more globally.
• Fields globalized by design require less current. In other words, global fields are more efficient in stimulating the proper structures, which cannot, therefore, be local to the basilar membrane.
• The efficiency of global stimulation strongly implies that the most probable site of stimulation is the spiral ganglion cell bodies in the modiolus.
• Current neurophysiologic theory cannot explain how a global stimulation can result in the perception of specific frequencies, nor can we easily imagine how and why the neural structures do not go into refractory.
• The geometry required by the tonotopic theory is not satisfied by either 6 mm or 25 mm electrodes. At best, the longer electrode could in theory only provide frequencies down to around 1500 Hz. If the theory were valid, the shorter electrodes could not offer patients access to frequencies below 4000 Hz. However, the audiograms of implant patients do not differ on the basis of the length of the electrode used, nor do they reflect these limits, again demonstrating that the tonotopic theory is invalid.
• It seems inescapable: When clinical results correspond so closely to anatomical investigation, the body of information resulting surely cannot be ignored, nor can it be successfully opposed merely because it contradicts an unsupported theory. It is time to admit that years of accumulating evidence have repeatedly shown that the theory of tonotopic stimulation in cochlear implants is simply wrong, and to place it on the dust heap of history.

This demise is really too bad: it was a lovely theory, and offered us the comforting illusion that we knew something. However, dead is dead. If we hold the truth higher than our own previous investments and current opinions, we need to recognize that the facts demonstrate that we were all wrong, and move on. This is called progress: it is rarely comfortable.

So what?

Of course, regardless of whether the tonotopic theory is invalid, some may ask why we should not have and use long electrodes anyway. After all, further information may somehow and someday show that there is some modest or subtle benefit to a tonotopic approach which we cannot currently discern, and, after all, the market is (as of this writing) filled primarily with long-electrode implants. Therefore even if the theory is wrong, so what? Why not use long (e.g. tonotopically correct) electrodes anyway?

In fact, this question is easy to answer.

Consider first that the category of patients for whom implants are seen as useful is rapidly broadening. For example, because of the uncertainties regarding the potential damage that the electrical fields which implants generate, implants were at one time not used in very young children. The questions regarding whether these electrical fields will damage a developing auditory system, however, are virtually resolved. We now know that electric field blanketing of the spiral ganglion cells do not cause damage, indeed, they appear to be beneficial in preventing atrophy.

There were uncertainties, as the Bilger report evidences, regarding whether cochlear implants could help patients detect and produce speech. These questions too have been resolved: we are now certain they will so assist.

One result of the gradual and now virtually complete resolution of these particular uncertainties — that is, we know they will not harm and will greatly assist — has been that cochlear implants are being seen by more and more clinicians as assistive devices useful both for younger patients, and for patients who are less than profoundly deaf. As such, many of the patients now being considered for implants, therefore, will have some remaining hair cells, and some residual hearing.

This brings up many possibilities. For example, one can visualize a time when implants might be used in combination with hearing aids to assist patients with frequency-limited losses, such as ski-drop high frequency losses, where these losses are caused by damage to the hair cells.

Long electrode damage

The widening of the group of those who are seen as suitable candidates and the fact of long electrodes however, have a tragic collision, as shown by the temporal bone studies of Linthicum, which reveal that long electrodes damage the cochlear structures.

The mechanism of the damage is apparently that as the electrode is inserted into the basal coil, at about the 7 mm depth it contacts and is constrained by the spiral ligament, which in turn is bounded by the bony canal of the cochlea itself. From that point, it is forced to turn inward and upward, following that spiral. (See Figure 5)

What appears to happen is that the electrode, which is forced thus to bend, presses against and strips off the spiral ligament, (See Figure 6) and damages the stria vascularis, the basilar membrane, and the organ of Corti. (See Figure 7)

Other studies, [30] [31] [32] have offered similar findings, and the matter now seems well established: The insertion of a long electrode damages and in some cases completely destroys cochlear structures which are essential to residual hearing. As a direct result, residual hearing is damaged or destroyed.

Schuknecht [33] in a discussion of degeneration of the cochlear neurons states:

Injury to the organ of Corti causes a retrograde neuronal degeneration proportionate to the extent of [the] loss of supporting cells.

In a personal communication, Schuknecht indicated that it was his estimate that 40% of the intact spiral ganglion cells would die if their dendrites were damaged. [34]

What evidence do we have regarding the damage which may or may not be associated with short electrodes? Case 580, case 591, and case 486 in the Linthicum study [35] show that a 6 mm electrode causes little or no damage. As such we conclude that any risk is greatly minimized if the electrode is inserted no deeper than 5 or 6 mm. Indeed, the evidence presented gives us reason to believe that all that is necessary is to simply insert the electrode; whether it is inserted 1 mm or 6 mm will not make any functional difference.

The remaining 10 bones studied by Linthicum, all of which had been implanted with long electrodes, each showed extensive damage to the spiral ligament and basilar membrane, and some even had fractures of the bone of the spiral limbus, testifying to the relative violence suffered by these very delicate structures.

Loss of Residual Hearing

As indicated, the anatomic damage from long electrodes is clinically manifested through the loss of pre-implant, unaided hearing. Bogies [36] and numerous clinicians, [37] and the more extensive study of severely hearing-impaired patients with Nucleus 22-channel implants done by the Cochlear Corporation, [38] have shown that residual hearing is likely to be lost as an apparent result of long electrode implantation.

By contrast, previous studies reported by Berliner et al [39] have shown that introduction of a 5 or 6 mm electrode into the scala tympani has very little chance of eliminating any residual hearing. On page 73 of the cited reference, they wrote:

The issue of possible damage to residual hearing has also been raised as a concern in the implantation of children. This has been evaluated by Dye et al. The records of 134 children implanted with the 3M/House single-channel cochlear implant at the House Ear Institute as of April 1987 were searched for any cases with measurable unaided pre-operative responses at or beyond 1000 Hz. Only 9 patients met this criterion of residual hearing, and post-implant data were available for 6 of them. A change in hearing was defined as greater than 5 dB. For 5 of the 6 children, no significant changes were observed at any frequency. One child had an increase of 6 dB at one frequency only (500 Hz) in both ears, post-implant. Thus, there is no evidence that the short 6 mm electrode used in the 3M/House implant will necessarily destroy residual hearing that might be present in a child receiving this implant.

In sum, the use of long electrodes in cochlear implants endangers and can often destroy the very capacity these implants are now meant to replace or which in the future they would supplement. This is the tragic irony: even if we wish to assert that there is, as yet, some subtle tonotopic effect which our tests have not revealed, current designs of long electrodes destroy the very structures they purport to stimulate.

The risks of surgery

One of the results of the space program is considerable information regarding the reliability of electronic assemblies. It is thus now well-known that each additional part added to such an assembly imposes an additional risk of failure. One practical consequence that we might expect is that complex internal hardware would be subject to a higher failure rate. This implies that our electrode designs must be as simple as possible, so that there are fewer parts to wear out or break, since whenever internal hardware fails, it must be explanted and a new electrode set implanted, necessitating surgery, with all its consequent risks.

I am not aware of any data which has been published on the failure rate of the Nucleus 22 channel device, but several years ago the failure rate of the 22 channel system was reported at 2%. At the meeting on children's cochlear implants held in New York February 4th and 5th of 1994, the failure rate was stated to have increased to 8%.

The risk of surgery is also imposed with complex internal hardware whenever this complexity is the result of a close coupling between the theory of the processor and the design of the internal hardware. New processing schemes, in that instance, would require new internal hardware. Because of the primitive state of our knowledge, today's best internal hardware designs must be those which would accept almost any conceivable signal over the 50-100 year life of the implant.

Inescapable conclusions

If long electrodes do not and cannot serve the purpose intended by the tonotopic theory of stimulation, there is no reason to use them. If, further, they cause damage to the cochlea and its structures, then there are important reasons not to use them. "First do no harm." Under what circumstances is it permissible to permanently destroy something so valuable as the sensation of hearing, when no demonstrated benefit results?

Of course, before this information was available, it could not be said that this destruction was culpable; but now we know, and this knowledge has removed our innocence and made us liable. The future, and indeed our patients, will judge us on that basis.

In fact, in spite of this knowledge, we may have to continue implanting long electrodes in implant patients for the time being because of the lack of immediate alternatives. And for those who are profoundly deaf, it should be clear any implant is the better of two options.

In sum, short electrodes are today's design of choice: they are neither complex nor strongly tied to any particular processing scheme, and they guard cochlear structures and residual hearing. Based on the evidence available to us, we can now assert that long electrode designs are yesterday's solution, and should be discarded for future designs.