個人化人工電子耳刺激策略與語音處理器

Abstract

研究顯示，人工電子耳只需6~9 個有效頻道，即可幫助電子耳使用者在安靜的環 境下聽得懂英語，但是卻無法在吵雜的環境中獲得同樣效果，其它如聆聽音樂、 聲調式語言 (tonal language) 特別是國語的辨識效果更是不足。而增加有效電子耳 頻道數目是提升這些效能的方法之一。 利用之前國科會計畫的基礎，我們建立了一個四電極電流導引技術 (FECSS)。在 未發表的臨床測試結果中，有一位電子耳配戴者在音調分辨測試中區分出高達300 個音調，這是目前全球商業產品中紀錄的三倍。 但是，通用模型的結果並無法滿足所有配戴者。所以需針對每一位配戴者建立一 個使用者專屬模型 (user specific model)。於是提出一個根據配戴者的電流生理資 料 (如生理阻抗或最低可聽到聲音的刺激電流強度(T level)與可接受最大刺激電 流強度(M level)) 來建立使用者專屬模型的方法。藉此可用來改善個別配戴者的四 電極電流導引刺激策略。 有了個別的刺激參數，即可進行靜態的音調分辨測試。同時，也需建立一個及時 的動態刺激策略。而為了能測試研發之高精確度電極刺激策略，我們需要一個可 依照需求調整內部程式的語音處理器，來針對所發展出來的電極刺激策略做測 試。最後如果時間允許，我們會將發展的電極刺激策略配合語音處理器做完整的 動態臨床試驗，來評估高精確度電極刺激策略對電子耳配戴者的改善程度。

Research shows that 6 ~ 9 effective channels are sufficient for cochlear implant (CI) users to listen and understand English in a completely quiet environment. However, it is not enough for listening to speech in noise, music and tonal language, particularly Mandarin Chinese, which are the last major frontiers in CI research. One accepted approach to improve the hearing performance is to increase the effective number of channel CI users can perceive. Unfortunately, there is no easy way to do that. In his previous NSC work, the principal investigator (PI) has focused on refining the cochlear implant computer model. He developed a four-electrode current steering scheme (FECSS), based on a generic computer model, which is much more complex than the two-electrode current steering scheme (TECSS) available commercially, and it led to a dramatic improvement in clinical performance. In unpublished clinical tests, a CI user was able to rank more than 300 different pitches in a pitch ranking test, which is almost three times the current best record by using commercial cochlear implants. Unfortunately, it is found that one generic computer model, no matter how accurate it is, it cannot be applied to all CI patients. A new computer model is needed for each CI patient, i.e. a patient specific computer model is needed. The PI is proposing a novel method to create a patient specific CI computer model based on electrophysiological measurement such as impedance, voltage, current, T level (threshold) and M level (most comfortable level). Once the CI user specific computer model is created, it can be used to develop a FECSS for that particular patient. With the patient specific computer model and FECSS, a patient specific FECSS will be created and tested on "static" pitch ranking test first. Simultaneously, we will develop this patient specific FECSS into a real time dynamic CI stimulating strategy. In order to implement and test this stimulating strategy in the clinical setting, access to the internal programming of the commercially available speech processor that is compatible with the patients’ headpiece is needed. Since commercial cochlear implants contain proprietary information, we need to develop a new CI research and clinical speech processor to implement and test this patient specific FECSS stimulating strategy. In order to test our user specific FECSS stimulating strategy on CI patients, the cochlear implant speech processor needs to be compatible with commercial implants, e.g. Advanced Bionics implant. We will develop a new research and clinical speech processor based on off the shelf digital signal processor (DSP) boards coupled with a custom designed transcutaneous transmission circuit to provide power and instructions to the implant electronics and CI electrodes inside the patient bodies. A new mapping software will be developed to map the CI patients. The speech processor will be subjected to electromagnetic compatibility (EMC) and related tests to satisfy all the pre-clinical test requirements for medical devices. Lastly, we will test the new stimulating strategy in the new CI research and clinical speech processor in clinical experiments if time permits.