COLLET Entry for the 2007 Humie award for Human Competitive Results Title: Results on the fitting of Cochlear Implants (1) the complete title of one (or more) paper(s) published in the open literature describing the work that the author claims describes a human-competitive result, P. Legrand, C. Bourgeois-République, V. Péan, E. Harboun-Cohen, J. Levy-Véhel, B. Frachet, E. Lutton, P. Collet, "Interactive Evolution for Cochlear Implants Fitting", to appear in the next special issue on Medical Applications of Genetic Programming and Evolable Machines, edited by Stephen Smith and Stefano Cagnoni. The paper has been accepted, and the attached paper is the final version that will be printed in the journal. Acknowledgement of this has been confirmed by Stephen Smith in the following mail, so that we could participate to Humies 2007: On Tue, 22 May 2007, Stephen L. Smith wrote: > Dear Pierre, > > It is certainly my opinion that the paper has now been unconditionally > accepted and is "in press" and am happy to make a statement to this effect. > > Best regards, > > Steve. (2) the name, complete physical mailing address, e-mail address, and phone number of EACH author of EACH paper, Name: Pierrick Legrand Physical address: IMB, Institut de Mathématiques de Bordeaux, UMR CNRS 5251. Université de Bordeaux 2, UFR sciences et modélisation, 146 rue Léo Saignat, 33076 Bordeaux cedex, France Email: legrand@sm.u-bordeaux2.fr Tel: +33 557 574 634 Name: Claire Bourgeois-République Physical address: Le2i, UMRS CNRS 5158, BP 47870, 21078 Dijon cedex - France Email: claire.bourgeois-republique@u-bourgogne.fr Tel : +33 380 395 912 Name: Vincent Péan Physical address: 1, promenade Jean Rostand, 93005 Bobigny cedex Email: vincent.pean@innotech.fr Tel: +33 149 469 495 Name: Esther Harboun-Cohen Physical address: Hôpital Avicenne (service ORL) 125 route de stalingrad 93000 Bobigny France Email: eharbouncohen@wanadoo.fr Tel: +33 146 910 765 Name: Jacques Levy-Vehel Physical address: INRIA Rocquencourt, B.P. 105 78153 LE CHESNAY Cedex, France Email: jacques.levy-vehel@irccyn.ec-nantes.fr Tel: +33 139 635 552 Name: Bruno Frachet Physical address: Hôpital Avicenne (service ORL) 125 route de stalingrad 93000 Bobigny France Email: bruno.frachet@avc.ap-hop-paris.fr Tel: +33 148 955 201 Name: Evelyne Lutton Physical address: INRIA Rocquencourt, B.P. 105 78153 LE CHESNAY Cedex, France Email: evelyne.lutton@inria.fr Tel: +33 139 635 523 Name: Pierre Collet Physical address: LIL-ULCO, BP 719, 62100 Calais cedex - France Email: pierre.collet@inria.fr Tel: +33 687 985 103 (3) the name of the corresponding author (i.e., the author to whom notices will be sent concerning the competition), Pierre Collet ( pierre.collet@inria.fr ) (4) the abstract of the paper(s), Cochlear implants are devices that become more and more sophisticated and adapted to the need of patients, but at the same time they become more and more difficult to parameterize. After a deaf patient has been surgically implanted, a specialised medical practitioner has to spend hours during months to precisely fit the implant to the patient. This process is a complex one implying two intertwined tasks: the practitioner has to tune the parameters of the device (optimisation) while the patient's brain needs to adapt to the new data he receives (learning). This paper presents a study that intends to make the implant more adaptable to environment (auditive ecology) and to simplify the process of fitting. Real experiments on volunteer implanted patients are presented, that show the efficiency of interactive evolution for this purpose. (5) a list containing one or more of the eight letters (A, B, C, D, E, F, G, or H) that correspond to the criteria (see above) that the author claims that the work satisfies, B, D, E, G, H (6) a statement stating why the result satisfies the criteria that the contestant claims (see the examples below as a guide to aid in constructing this part of the submission), Cochlear Implants are devices that turn acoustic sounds into electrical impulses via a Digital Signal Processor. The aim is to directly stimulate the auditory nerve through the cochlea so that deaf patients can "hear" sounds. The problem to solve is the following : how should the auditory area of the brain be stimulated (through the auditory nerve) in order to reproduce complex sounds such as the human voice, precisely enough so that deaf patients can understand speech without any visual assistance such as lip-reading (the most difficult test being for the patient to follow a conversation on the phone, for instance). Early cochlear implants only featured one electrode, so although they allowed deaf patients to hear sounds in their environment, they did not allow the patients to understand speech. Then, multi-electrode implants came out, and surprisingly enough some lucky patients were able to understand words with implants having 2 or 3 electrodes only. Nowadays, cochlear implants can drive as many as 15 or even 22 electrodes, each mapped onto different sound frequences and most patients are able to follow a conversation, even in a noisy environment like a restaurant. However, for some of them, the cochlear implant is not that miraculous, and some still find it difficult to understand basic speech. Some unfortunate patients cannot even make anything out of the strange metallic garble they hear. In such cases, some of them simply prefer to switch off the implant and remain deaf, while others do keep it on, just to be aware that something is happening in their vicinity. Right now, all cochlear implant manufacturers advocate to maximise the [T,C] interval for each of the 15 to 22 electrodes, in order to provide the auditory nerve with the maximum information, the T value being (for a specific electrode) the Threshold intensity below which the patient does not "hear" anything, and the C value being the "Comfort" value, above which intensity the stimulation becomes uncomfortable, and possibly dangerous for the auditory neurons facing the electrode. Then, expert practitioners and orthophonists say that a fitting can only be precisely evaluated after several days, by extensive tests taking around one hour to perform, meaning that in practice, only 2 or 3 fittings are tested during one fitting session. The patient is then sent back home for several weeks, with the possibility to revert to the older setting until he comes back to hospital, with 3 to a maximum of 10 fitting sessions a year. The aim of the work presented here was to provide a tool that could help expert practitioners find some better parameter settings (fittings) for patients who still cannot follow a conversation after they have received a cochlear implant. An interactive evolutionary algorithm has been devised to determine (among all the tunable parameters) good [T,C] intervals for each electrode of the implant, using a reduced evaluation procedure taking less than 4 minutes to complete. On the first tested patient, the new evaluation procedure (counting the number of recognised words in a set of calibrated sentences + taking into account the "comfort" of the fitting) gave a value of 48.5/100 for the best fitting obtained after 10 years with an expert practitioner, with less than 50% of understood words, and a "comfort" value of 5/10. After 89 evaluations made in one and a half day, the best individual obtained an evaluation of 91.5/100, but with a poor "comfort" value. The patient was sent back home with the previous setting, and one month later, a set of tests showed that results were roughly the same (showing that the reduced evaluation procedure gave reproductible values). The patient decided to go back home with the new fitting, and after a while, the fitting became comfortable, thanks to neural plasticity. A study of the evolved [T,C] intervals for the electrodes showed that the evolutionary algorithm effectively switched off all electrodes but 3 (out of 13 functional electrodes for this patient), and deterministic tests showed that for this particular patient, activating electrode 4 (a perfectly functional electrode otherwise) dramatically reduced speech comprehension. The conclusion was that even perfectly functional electrodes may not be beneficial to hearing, making it a combinatorial problem: "For a specific patient, which combination of electodes will maximise his hearing / speech understanding ?" The problem is that, with 22 electrodes and no way to tell whether a functional electrode is beneficial or not, the practitioner must test 2^22 = 4 billion combinations. The other conclusion was that although probably imprecise, a reduced and quick evaluation proved accurate enough to guide an evolutionary algorithm towards a maximisation of hearing performance. Tests were performed on a set of other patients who unfortunately could not even understand the words of the calibrated sentences used in the previous test. Other evaluation procedures were elaborated by orthophonists, based on the recognition of consonants and elementary speech sounds. Although it was not possible to obtain a large enough number of evaluations due to the length and complexity of the new evaluation, it was found that for all of them, fittings that removed several electrodes were found to be at least as good, or even better than manual fittings that were maximising the [T,C] interval for all electrodes. We therefore claim that the presented work fulfills criteria B, D, E, G, H for the following reasons: (B) The result is equal to or better than a result that was accepted as a new scientific result at the time when it was published in a peer-reviewed scientific journal: Obtained results (when the algorithm had enough evaluations to start working) were always equal or better than the current method of maximising [T,C] intervals for all electrodes, used and advocated by all cochlear implant manufacturers and experts. Until now, no peer-reviewed scientific journals have published another successful method than maximising all [T,C] intervals (mainly because it seems unreasonable to fitting experts to discard functional electrodes), which the findings of the presented work prove wrong, at least for some patients. (D) The result is publishable in its own right as a new scientific result - independent of the fact that the result was mechanically created: The results have been published in international medical symposia, cf. for instance: E. Harboun Cohen, V. Péan, C. Bourgeois République, P. Legrand, P. Collet, B. Philippon, M-C. Ouayoun, B. Frachet, ``Preliminary Results on Automatic Cochlear Implant Fitting Using an Interactive Evolutionary Algorithm,'' 9th International Conference on Cochlear implants and related Sciences, Vienna, Austria, June 2006. (E) The result is equal to or better than the most recent human-created solution to a long-standing problem for which there has been a succession of increasingly better human-created solutions: Cochlear implants have been around for more than 40 years and until now, all fitting experts were convinced that it did not make sense to deactivate some electrodes to enhance audition and/or speech understanding. (G) The result solves a problem of indisputable difficulty in its field: Finding the correct way to communicate directly with the brains by exciting in a deterministic way the neurons of the auditory nerve is indisputably one of the most difficult and challenging tasks a computer scientist could face. (H) The result holds its own or wins a regulated competition involving human contestants (in the form of either live human players or human-written computer programs). Although the problem to solve is not a game, there has been live human contestants (expert practitioners) trying to solve this problem for the last 40 years, and in the different tests, it has always been possible to find a better or equal solution than the one suggested by the expert. (7) a full citation of the paper (that is, author names; publication date; name of journal, conference, technical report, thesis, book, or book chapter; name of editors, if applicable, of the journal or edited book; publisher name; publisher city; page numbers, if applicable); Pierrick Legrand, Claire Bourgeois-République, Vincent Péan, Esther Harboun-Cohen, Jacques Levy-Vehel, Bruno Frachet, Evelyne Lutton, Pierre Collet, "Interactive Evolution for Cochlear Implants Fitting", Genetic Programming and Evolvable Machines, accepted in 2007. (8) a statement either that "any prize money, if any, is to be divided equally among the co-authors" OR a specific percentage breakdown as to how the prize money, if any, is to be divided among the co-authors. Any prize money, if any, will be divided equally among the co-authors. (9) a statement stating why the judges should consider the entry as "best" in comparison to other entries that may also be "human-competitive." Thanks to the presented interactive evolutionary algorithm, we have shown that it is possible to improve hearing by de-activating a number of functional electrodes, and that a quick reduced evaluation procedure could be sufficient to guide an optimisation algorithm. Such findings may open new directions that were not envisaged before. The reasons why this entry could be considered as "best" is that these results may drastically improve the social life of many totally deaf patients, who have undergone an expensive and potentially dangerous surgical intervention, to be implanted with a device that they cannot profitably use.