MyOOPS開放式課程
請加入會員以使用更多個人化功能
來自全球頂尖大學的開放式課程,現在由世界各國的數千名義工志工為您翻譯成中文。請免費享用!
課程來源:TED
     

 

Sheila Nirenberg 談治療失明的人工眼

Sheila Nirenberg: A prosthetic eye to treat blindness

 

Photo of three lions hunting on the Serengeti.

講者:Sheila Nirenberg

2011年11月演講,2012年1月在TEDMED 2011上線

 

翻譯:TED

編輯:朱學恆、洪曉慧

簡繁轉換:洪曉慧

後製:洪曉慧

字幕影片後制:謝旻均

 

影片請按此下載

MAC及手持裝置版本請按此下載

閱讀中文字幕純文字版本

 

關於這場演講

在TEDMED中,Sheila Nirenberg展示一個使視障人士能接收到影像的大膽方式:藉由一個連接視神經的鏡頭,直接將訊號傳到大腦中。

 

關於Sheila Nirenberg

Sheila Nirenberg研究大腦如何將訊息編碼-也許我們能將其解碼,藉此開發人工感應裝置。

 

為什麼要聽她演講

Sheila Nirenberg是神經科學家及康乃爾大學威爾醫學院教授,研究神經編碼-也就是大腦如何從外界接收資訊,並將它編碼成電活動模式。這個研究的想法是將電流訊號解碼成電脈衝模式,藉此得知動物所見、所思或感覺。她最近致力於利用這項研究開發新型人工裝置,尤其是治療失明的裝置。

 

Sheila Nirenberg的英語網上資料

 

[TED科技‧娛樂‧設計]

已有中譯字幕的TED影片目錄(繁體)(簡體)。請注意繁簡目錄是不一樣的。

 

Sheila Nirenberg 談治療失明的人工眼

 

我研究的是大腦如何處理資訊,也就是大腦如何接受外界資訊,並將它轉換成腦電波模式,及如何利用這些模式來控制行為-例如看、聽和拿東西。我是基礎科學家,而非臨床科學家,但過去一年半裡,我開始轉換跑道,利用對腦電波模式的瞭解來開發人工眼裝置。我今天想向各位展示其中一個例子,這是我們在此項研究中第一個嘗試-開發一種治療失明的人工眼裝置。

 

我先從失明這個問題談起。美國的失明者有一千萬人,全球因視網膜疾病或黃斑部病變而失明,或面臨失明危機的人數更多,但我們卻對此束手無策。雖然有一些治療性藥物,但只對少部分患者有效。因此對大部分病人來說,他們恢復視力的希望得寄託於人工眼設備。問題是,目前人工眼的效果並不好,它們提供的視覺範圍依然有限。因此,例如,這些裝置只能幫助病人看到亮光及高對比度的輪廓,幾乎只有這樣而已,所以跟正常視覺所見的影像相距甚遠。

 

今天我想告訴諸位的是,我們正在研發的一項設備很可能使這個情況改觀,使病人擁有更有效的視覺能力,我想向諸位展示一下它的運作原理。好,首先我稍微講一點背景知識,說明正常視網膜的運作原理,讓大家明白我們要解決的問題是什麼。這是一個視網膜,所以這是視網膜和大腦的圖片。當你看著某樣物體,例如這張嬰兒臉部圖片,圖像進入眼睛,落在視網膜上。它落在視網膜的前端細胞,即光感受器上,然後中間部分,即視網膜的電路系統,開始處理這個圖像。它對圖像進行操作,從中提取資訊,然後將資訊轉換成代碼,代碼以電脈衝模式傳入大腦,所以其中關鍵就是將圖像轉換成代碼。我所說的代碼,就是字面上的意思,像這個脈衝模式代表的就是「嬰兒臉」,所以當大腦得到這個脈衝模式,就知道這代表嬰兒臉孔,如果它得到不同模式,就知道代表的是什麼物體,例如一隻狗或另一種代表房子的模式,總之就是這麼回事。

 

當然,現實生活是動態的,意味著外界景物總是不斷變化,所以脈衝模式也會不斷變化,因為你看到的世界也不斷地變化著。所以,你知道,這就有點複雜了。你的眼睛每毫秒都會傳送不同的脈衝模式,告訴大腦你看到的事物,所以當某人罹患視網膜退化疾病,例如黃斑部病變時,會發生什麼事?會發生的情況是,前端細胞會死亡,光感受器會死亡,接著,所有連接這些組織的電路和細胞都會停止作用,最後剩下的只有那些輸出細胞,就是將訊號傳入大腦的細胞。但因為這些病變,它們無法得到輸入訊號,所以也無法輸出訊號,使大腦無法再獲得任何視覺資訊,也就是說,他(她)失明了。

 

所以,一個解決方案是,製造一種設備來模仿前端細胞的電路系統功能,將訊號傳給視網膜的輸出細胞,使它們能重新開始正常運作,將訊號輸入大腦,所以這就是我們研究的方向,也是我們人工眼的功能。所以它由兩個部分組成,我們稱之為編碼器和感測器。編碼器就像我之前提過的-可模仿前端細胞的電路功能-所以它可接收圖像,並轉換成視網膜能接受的編碼,然後感測器使輸出細胞將編碼送進大腦,這樣視網膜人工眼就能產生正常的視網膜輸出訊號。因此一個完全失去作用的視網膜,即使前端細胞完全失去電路功能,沒有光感受器,現在也能輸出正常訊號,一種大腦能理解的訊號,目前還沒有其他設備能做到這一點。

 

好,我簡單說明一下編碼器的功能,因為這是相當關鍵的部分,很有意思也很酷。我不確定是否該用「酷」來形容,不過就是那個意思。它的功能是,用一組方程式替代最重要的視網膜電路系統部分,一組可植入晶片的方程式,所以,這其實就是數學,也就是說,我們並非真的換掉視網膜的組成部份,並非我們做了個迷你設備來替代每個不同類型的細胞,我們只是用一組方程式來擷取視網膜的功能,所以,以某種程度來說,這些方程式類似於編碼簿,一個圖像進入,通過這組方程式,變成一串電脈衝輸出,就像正常視網膜所產生的一樣。

 

現在,我實際展示一下,我們確實能產生正常的輸出訊號,以及這麼做的意義。這是三組放電模式,最上面的來自視力正常的動物,中間的來自使用這種編碼感測設備的失明動物,底部的來自使用一般人工眼的失明動物。所以底部是目前最尖端的科技設備,主要由光感測器組成,但沒有編碼器。我們所做的就是把日常事物-人、嬰兒、公園座椅等這些一般物體的影片給牠們看,然後記錄三組動物的視網膜反應。我說明一下,每個長方形代表幾個細胞的放電模式,就像之前投影片中所顯示的一樣,每一行代表一個不同的細胞,我將這些脈衝顯示得更纖細些,讓你們能看見更詳細的資料。

 

好,如圖所示,使用編碼感測設備的失明動物,放電模式跟正常放電模式相當接近-雖不完美,但已經很好了-使用一般人工眼的失明動物顯示效果相當差;使用一般人工眼時,細胞確實可以放電,只是並非以正常放電模式呈現,因為它們沒有正確的編碼。編碼有多重要?它對病人視力的潛在影響有多大?我向各位展示一個基礎實驗來回答這個問題。當然,我還有許多其他資料,如果你有興趣,我很樂意多展示一些。所以這個實驗被稱為重建實驗,所以我們從記錄裡截取某一刻,觀察那一刻視網膜看到了什麼?我們是否能重建視網膜從這些放電模式中看到的東西?

 

所以我們分別觀察一般人工眼和編碼感測器的反應,我先從一般人工眼開始,讓你們看一下實驗結果。所以,你們可以看出,效果很有限,因為這個放電模式沒有經過正確編碼,很難使人看出外界的影像。你可以看見那裡有些東西,但看不清楚是什麼,這又回到我開頭時所說的問題,它使病人能看見高對比度的輪廓和亮光,但很難達到更佳效果。好,這是什麼圖像?嬰兒的臉,那麼,使用我們的方法加入編碼後,會看見什麼?你們可以看出,效果好多了,你不僅能看出這是一張嬰兒臉,也能看出是這個嬰兒的臉。這確實是相當具挑戰性的任務,所以左邊是單用編碼器的結果,右邊是將它使用於失明的視網膜,所以是編碼器加上感測器,但具有關鍵作用的其實是編碼器,因為我們可以將編碼器與不同的感測器結合起來。

 

這僅是我們所做的第一個嘗試。我想談一些關於一般人工眼的問題。它剛剛面世時確實令人振奮,因為這畢竟是能讓失明視網膜產生反應的方法,但其中的限制因素在於編碼,如何能讓細胞產生更好的反應,產生正常的反應,所以,這就是我們所做的貢獻。現在我總結一下。如我之前所說的,如果你有興趣,我還有很多其他資料,我現在只簡單講述一些基本概念,這種能用自己的語言與大腦溝通的方法,以及這種方法的潛在力量。所以這跟運動修復術不同,它是將大腦的指令傳入一個設備,我們則是將外界資訊輸入大腦,並使其被大腦理解。

 

最後我想強調的是,這個方法是廣泛適用的,我們尋找視網膜編碼的技術也可用於尋找人體其他區域的編碼,例如聽覺系統和運動系統,以治療聽力和運動障礙。使用跟我們繞過受損視網膜電路系統,進入視網膜輸出細胞相同的方法,我們也可繞過受損的耳蝸電路系統,直達聽神經,或繞過大腦皮質中受損的區域,進入運動皮質區,彌補中風造成的損害。

 

最後,我想給大家一個簡單的概念。瞭解代碼非常、非常重要,如果我們能瞭解代碼,也就是大腦的語言,一切之前看似無法做到的事都變得可能。謝謝。

 

(掌聲)

 

以下為系統擷取之英文原文

About this Talk

At TEDMED, Sheila Nirenberg shows a bold way to create sight in people with certain kinds of blindness: by hooking into the optic nerve and sending signals from a camera direct to the brain.

About the Speaker

Sheila Nirenberg studies how the brain encodes information -- possibly allowing us to decode it, and maybe develop prosthetic sensory devices. Full bio »

Transcript

I study how the brain processes information. That is, how it takes information in from the outside world, and converts it into patterns of electrical activity, and then how it uses those patterns to allow you to do things -- to see, hear, to reach for an object. So I'm really a basic scientist, not a clinician, but in the last year and a half I've started to switch over, to use what we've been learning about these patterns of activity to develop prosthetic devices, and what I wanted to do today is show you an example of this. It's really our first foray into this. It's the development of a prosthetic device for treating blindness.

So let me start in on that problem. There are 10 million people in the U.S. and many more worldwide who are blind or are facing blindness due to diseases of the retina, diseases like macular degeneration, and there's little that can be done for them. There are some drug treatments, but they're only effective on a small fraction of the population. And so, for the vast majority of patients, their best hope for regaining sight is through prosthetic devices. The problem is that current prosthetics don't work very well. They're still verylimited in the vision that they can provide. And so, you know, for example, with thesedevices, patients can see simple things like bright lights and high contrast edges, not very much more, so nothing close to normal vision has been possible.

So what I'm going to tell you about today is a device that we've been working on that I think has the potential to make a difference, to be much more effective, and what I wanted to do is show you how it works. Okay, so let me back up a little bit and show you how a normal retina works first so you can see the problem that we were trying to solve. Here you have a retina. So you have an image, a retina, and a brain. So when you look at something, like this image of this baby's face, it goes into your eye and it lands on your retina, on the front-end cells here, the photoreceptors. Then what happens is the retinal circuitry, the middle part, goes to work on it, and what it does is it performs operations on it, it extracts information from it, and it converts that information into a code. And the code is in the form of these patterns of electrical pulses that get sent up to the brain, and so the key thing isthat the image ultimately gets converted into a code. And when I say code, I do literally mean code. Like this pattern of pulses here actually means "baby's face," and so when the brain gets this pattern of pulses, it knows that what was out there was a baby's face, and if it got a different pattern it would know that what was out there was, say, a dog, or another pattern would be a house. Anyway, you get the idea.

And, of course, in real life, it's all dynamic, meaning that it's changing all the time, so the patterns of pulses are changing all the time because the world you're looking at is changing all the time too. So, you know, it's sort of a complicated thing. You have these patterns of pulses coming out of your eye every millisecond telling your brain what it is that you're seeing. So what happens when a person gets a retinal degenerative disease likemacular degeneration? What happens is is that, the front-end cells die, the photoreceptors die, and over time, all the cells and the circuits that are connected to them, they die too. Until the only things that you have left are these cells here, the output cells,the ones that send the signals to the brain, but because of all that degeneration they aren't sending any signals anymore. They aren't getting any input, so the person's brain no longer gets any visual information -- that is, he or she is blind.

So, a solution to the problem, then, would be to build a device that could mimic the actions of that front-end circuitry and send signals to the retina's output cells, and they can go back to doing their normal job of sending signals to the brain. So this is what we've been working on, and this is what our prosthetic does. So it consists of two parts, what we call an encoder and a transducer. And so the encoder does just what I was saying: it mimics the actions of the front-end circuitry -- so it takes images in and converts them into the retina's code. And then the transducer then makes the output cells send the code on up to the brain, and the result is a retinal prosthetic that can produce normal retinal output.So a completely blind retina, even one with no front-end circuitry at all, no photoreceptors,can now send out normal signals, signals that the brain can understand. So no other device has been able to do this.

Okay, so I just want to take a sentence or two to say something about the encoder and what it's doing, because it's really the key part and it's sort of interesting and kind of cool.I'm not sure "cool" is really the right word, but you know what I mean. So what it's doing is, it's replacing the retinal circuitry, really the guts of the retinal circuitry, with a set of equations, a set of equations that we can implement on a chip. So it's just math. In other words, we're not literally replacing the components of the retina. It's not like we're making a little mini-device for each of the different cell types. We've just abstracted what theretina's doing with a set of equations. And so, in a way, the equations are serving as sort of a codebook. An image comes in, goes through the set of equations, and out comes streams of electrical pulses, just like a normal retina would produce.

Now let me put my money where my mouth is and show you that we can actually produce normal output, and what the implications of this are. Here are three sets of firing patterns. The top one is from a normal animal, the middle one is from a blind animal that's been treated with this encoder-transducer device, and the bottom one is from a blind animal treated with a standard prosthetic. So the bottom one is the state-of-the-art device that's out there right now, which is basically made up of light detectors, but no encoder. So what we did was we presented movies of everyday things -- people, babies, park benches, you know, regular things happening -- and we recorded the responses from the retinas of these three groups of animals. Now just to orient you, each box is showing the firing patterns of several cells, and just as in the previous slides, each row is a different cell,and I just made the pulses a little bit smaller and thinner so I could show you a long stretch of data.

So as you can see, the firing patterns from the blind animal treated with the encoder-transducer really do very closely match the normal firing patterns -- and it's not perfect, but it's pretty good -- and the blind animal treated with the standard prosthetic, the responses really don't. And so with the standard method, the cells do fire, they just don't fire in the normal firing patterns because they don't have the right code. How important is this?What's the potential impact on a patient's ability to see? So I'm just going to show you onebottom-line experiment that answers this, and of course I've got a lot of other data, so if you're interested I'm happy to show more. So the experiment is called a reconstruction experiment. So what we did is we took a moment in time from these recordings and asked, what was the retina seeing at that moment? Can we reconstruct what the retinawas seeing from the responses from the firing patterns?

So, when we did this for responses from the standard method and from our encoder and transducer. So let me show you, and I'm going to start with the standard method first. So you can see that it's pretty limited, and because the firing patterns aren't in the right code, they're very limited in what they can tell you about what's out there. So you can see thatthere's something there, but it's not so clear what that something is, and this just sort ofcircles back to what I was saying in the beginning, that with the standard method, patients can see high-contrast edges, they can see light, but it doesn't easily go further than that. So what was the image? It was a baby's face. So what about with our approach, adding the code? And you can see that it's much better. Not only can you tell that it's a baby's face, but you can tell that it's this baby's face, which is a really challenging task. So on the left is the encoder alone, and on the right is from an actual blind retina, so the encoder and the transducer. But the key one really is the encoder alone, because we can team up the encoder with the different transducer.

This is just actually the first one that we tried. I just wanted to say something about the standard method. When this first came out, it was just a really exciting thing, the idea that you even make a blind retina respond at all. But there was this limiting factor, the issue of the code, and how to make the cells respond better, produce normal responses, and so this was our contribution. Now I just want to wrap up, and as I was mentioning earlier of course I have a lot of other data if you're interested, but I just wanted to give this sort of basic idea of being able to communicate with the brain in its language, and the potential power of being able to do that. So it's different from the motor prosthetics where you're communicating from the brain to a device. Here we have to communicate from the outside world into the brain and be understood, and be understood by the brain.

And then the last thing I wanted to say, really, is to emphasize that the idea generalizes.So the same strategy that we used to find the code for the retina we can also use to find the code for other areas, for example, the auditory system and the motor system, so for treating deafness and for motor disorders. So just the same way that we were able tojump over the damaged circuitry in the retina to get to the retina's output cells, we can jump over the damaged circuitry in the cochlea to get the auditory nerve, or jump over damaged areas in the cortex, in the motor cortex, to bridge the gap produced by a stroke.

I just want to end with a simple message that understanding the code is really, really important, and if we can understand the code, the language of the brain, things becomepossible that didn't seem obviously possible before. Thank you.

(Applause)


留下您對本課程的評論
標題:
您目前為非會員,留言名稱將顯示「匿名非會員」
只能進行20字留言

留言內容:

驗證碼請輸入3 + 1 =

標籤

現有標籤:1
新增標籤:


有關本課程的討論

目前暫無評論,快來留言吧!

Creative Commons授權條款 本站一切著作係採用 Creative Commons 授權條款授權。
協助推廣單位: