Research press release


Nature Communications

Neuroscience: Decoding brain activity to spell sentences

発話できない臨床試験参加者が、言葉を発さずに音声学的に単語を綴って、完全な文を書こうとする時の脳活動を解読するデバイスについて報告する論文が、Nature Communicationsに掲載される。この知見は、発話のための無音制御型神経プロテーゼが、単語を綴る行為に基づいた方法で、文章を生成する可能性を浮き彫りにしている。


今回、Edward Changたちは、脳の活動を1文字に翻訳してリアルタイムで完全な文を綴ることができる神経プロテーゼを設計し、重度の声道麻痺と四肢麻痺のために意思伝達能力に障害を来した試験参加者に使用させて、この神経プロテーゼの有用性を実証した。この研究でChangたちは、フォネティックコードに関連する脳活動を解読するシステムを設計することで、過去の研究で用いた手法をさらに大規模な語彙に拡大適用した。このシステムの検証試験では、試験参加者が無言でフォネティックコードの発声を試みた時の脳活動の解読に成功し、1,152語の語彙から毎分29.4文字の速度で文章が生成され、文字の平均誤読率は6.13%だった。さらなる実験で、この手法は9,000語以上を含む大規模な語彙に一般化され、平均誤読率は8.23%となった。


A device capable of decoding brain activity in a participant with speech paralysis, as they silently attempt to spell out words phonetically to create full sentences, is reported in Nature Communications. The findings highlight the potential of a silently controlled speech neuroprosthesis to generate sentences through a spelling-based approach.

Neuroprostheses are devices that replace lost nervous system function, and have the potential to restore communication to people who cannot speak or type due to paralysis. However, it is unclear if silent attempts to speak can be used to control a communication neuroprosthesis. Previous research has shown that a neuroprosthetic system in a participant with speech paralysis can decode up to 50 words. However, this system was limited to a specific vocabulary and the participant had to attempt to speak the words out loud, which required significant effort given their paralysis.

Edward Chang and colleagues designed a neuroprosthesis capable of translating brain activity into single letters to spell out full sentences in real time, and demonstrated its use in a participant who suffered from limited communication because of severe vocal and limb paralysis. The authors expanded the previous approach to a larger vocabulary by designing their system to decode brain activity associated with the phonetic alphabet. In tests, the device was able to decode the brain activity of the participant as they attempted to silently speak each letter phonetically to produce sentences from a 1,152-word vocabulary at a speed of 29.4 characters per minute, and an average character error rate of 6.13%. In further experiments, the authors found that the approach generalized to large vocabularies containing over 9,000 words, averaging a 8.23% error rate.

The results highlight the potential of silently controlled speech neuroprostheses to generate sentences through a spelling-based approach using phonetic code words. Further work is required to demonstrate if this approach is reproducible in more participants.

doi: 10.1038/s41467-022-33611-3


メールマガジンリストの「Nature 関連誌今週のハイライト」にチェックをいれていただきますと、毎週各ジャーナルからの最新の「注目のハイライト」をまとめて皆様にお届けいたします。