Sheila Nirenberg: A prosthetic eye to treat blindness

100,610 views ・ 2011-12-20

TED


Please double-click on the English subtitles below to play the video.

00:15
I study how the brain processes
0
15330
2000
00:17
information. That is, how it takes
1
17330
2000
00:19
information in from the outside world, and
2
19330
2000
00:21
converts it into patterns of electrical activity,
3
21330
2000
00:23
and then how it uses those patterns
4
23330
2000
00:25
to allow you to do things --
5
25330
2000
00:27
to see, hear, to reach for an object.
6
27330
2000
00:29
So I'm really a basic scientist, not
7
29330
2000
00:31
a clinician, but in the last year and a half
8
31330
2000
00:33
I've started to switch over, to use what
9
33330
2000
00:35
we've been learning about these patterns
10
35330
2000
00:37
of activity to develop prosthetic devices,
11
37330
3000
00:40
and what I wanted to do today is show you
12
40330
2000
00:42
an example of this.
13
42330
2000
00:44
It's really our first foray into this.
14
44330
2000
00:46
It's the development of a prosthetic device
15
46330
2000
00:48
for treating blindness.
16
48330
2000
00:50
So let me start in on that problem.
17
50330
2000
00:52
There are 10 million people in the U.S.
18
52330
2000
00:54
and many more worldwide who are blind
19
54330
2000
00:56
or are facing blindness due to diseases
20
56330
2000
00:58
of the retina, diseases like
21
58330
2000
01:00
macular degeneration, and there's little
22
60330
2000
01:02
that can be done for them.
23
62330
2000
01:04
There are some drug treatments, but
24
64330
2000
01:06
they're only effective on a small fraction
25
66330
2000
01:08
of the population. And so, for the vast
26
68330
2000
01:10
majority of patients, their best hope for
27
70330
2000
01:12
regaining sight is through prosthetic devices.
28
72330
2000
01:14
The problem is that current prosthetics
29
74330
2000
01:16
don't work very well. They're still very
30
76330
2000
01:18
limited in the vision that they can provide.
31
78330
2000
01:20
And so, you know, for example, with these
32
80330
2000
01:22
devices, patients can see simple things
33
82330
2000
01:24
like bright lights and high contrast edges,
34
84330
2000
01:26
not very much more, so nothing close
35
86330
2000
01:28
to normal vision has been possible.
36
88330
3000
01:31
So what I'm going to tell you about today
37
91330
2000
01:33
is a device that we've been working on
38
93330
2000
01:35
that I think has the potential to make
39
95330
2000
01:37
a difference, to be much more effective,
40
97330
2000
01:39
and what I wanted to do is show you
41
99330
2000
01:41
how it works. Okay, so let me back up a
42
101330
2000
01:43
little bit and show you how a normal retina
43
103330
2000
01:45
works first so you can see the problem
44
105330
2000
01:47
that we were trying to solve.
45
107330
2000
01:49
Here you have a retina.
46
109330
2000
01:51
So you have an image, a retina, and a brain.
47
111330
2000
01:53
So when you look at something, like this image
48
113330
2000
01:55
of this baby's face, it goes into your eye
49
115330
2000
01:57
and it lands on your retina, on the front-end
50
117330
2000
01:59
cells here, the photoreceptors.
51
119330
2000
02:01
Then what happens is the retinal circuitry,
52
121330
2000
02:03
the middle part, goes to work on it,
53
123330
2000
02:05
and what it does is it performs operations
54
125330
2000
02:07
on it, it extracts information from it, and it
55
127330
2000
02:09
converts that information into a code.
56
129330
2000
02:11
And the code is in the form of these patterns
57
131330
2000
02:13
of electrical pulses that get sent
58
133330
2000
02:15
up to the brain, and so the key thing is
59
135330
2000
02:17
that the image ultimately gets converted
60
137330
2000
02:19
into a code. And when I say code,
61
139330
2000
02:21
I do literally mean code.
62
141330
2000
02:23
Like this pattern of pulses here actually means "baby's face,"
63
143330
3000
02:26
and so when the brain gets this pattern
64
146330
2000
02:28
of pulses, it knows that what was out there
65
148330
2000
02:30
was a baby's face, and if it
66
150330
2000
02:32
got a different pattern it would know
67
152330
2000
02:34
that what was out there was, say, a dog,
68
154330
2000
02:36
or another pattern would be a house.
69
156330
2000
02:38
Anyway, you get the idea.
70
158330
2000
02:40
And, of course, in real life, it's all dynamic,
71
160330
2000
02:42
meaning that it's changing all the time,
72
162330
2000
02:44
so the patterns of pulses are changing
73
164330
2000
02:46
all the time because the world you're
74
166330
2000
02:48
looking at is changing all the time too.
75
168330
3000
02:51
So, you know, it's sort of a complicated
76
171330
2000
02:53
thing. You have these patterns of pulses
77
173330
2000
02:55
coming out of your eye every millisecond
78
175330
2000
02:57
telling your brain what it is that you're seeing.
79
177330
2000
02:59
So what happens when a person
80
179330
2000
03:01
gets a retinal degenerative disease like
81
181330
2000
03:03
macular degeneration? What happens is
82
183330
2000
03:05
is that, the front-end cells die,
83
185330
2000
03:07
the photoreceptors die, and over time,
84
187330
2000
03:09
all the cells and the circuits that are
85
189330
2000
03:11
connected to them, they die too.
86
191330
2000
03:13
Until the only things that you have left
87
193330
2000
03:15
are these cells here, the output cells,
88
195330
2000
03:17
the ones that send the signals to the brain,
89
197330
2000
03:19
but because of all that degeneration
90
199330
2000
03:21
they aren't sending any signals anymore.
91
201330
2000
03:23
They aren't getting any input, so
92
203330
2000
03:25
the person's brain no longer gets
93
205330
2000
03:27
any visual information --
94
207330
2000
03:29
that is, he or she is blind.
95
209330
3000
03:32
So, a solution to the problem, then,
96
212330
2000
03:34
would be to build a device that could mimic
97
214330
2000
03:36
the actions of that front-end circuitry
98
216330
2000
03:38
and send signals to the retina's output cells,
99
218330
2000
03:40
and they can go back to doing their
100
220330
2000
03:42
normal job of sending signals to the brain.
101
222330
2000
03:44
So this is what we've been working on,
102
224330
2000
03:46
and this is what our prosthetic does.
103
226330
2000
03:48
So it consists of two parts, what we call
104
228330
2000
03:50
an encoder and a transducer.
105
230330
2000
03:52
And so the encoder does just
106
232330
2000
03:54
what I was saying: it mimics the actions
107
234330
2000
03:56
of the front-end circuitry -- so it takes images
108
236330
2000
03:58
in and converts them into the retina's code.
109
238330
2000
04:00
And then the transducer then makes the
110
240330
2000
04:02
output cells send the code on up
111
242330
2000
04:04
to the brain, and the result is
112
244330
2000
04:06
a retinal prosthetic that can produce
113
246330
3000
04:09
normal retinal output.
114
249330
2000
04:11
So a completely blind retina,
115
251330
2000
04:13
even one with no front-end circuitry at all,
116
253330
2000
04:15
no photoreceptors,
117
255330
2000
04:17
can now send out normal signals,
118
257330
2000
04:19
signals that the brain can understand.
119
259330
3000
04:22
So no other device has been able
120
262330
2000
04:24
to do this.
121
264330
2000
04:26
Okay, so I just want to take
122
266330
2000
04:28
a sentence or two to say something about
123
268330
2000
04:30
the encoder and what it's doing, because
124
270330
2000
04:32
it's really the key part and it's
125
272330
2000
04:34
sort of interesting and kind of cool.
126
274330
2000
04:36
I'm not sure "cool" is really the right word, but
127
276330
2000
04:38
you know what I mean.
128
278330
2000
04:40
So what it's doing is, it's replacing
129
280330
2000
04:42
the retinal circuitry, really the guts of
130
282330
2000
04:44
the retinal circuitry, with a set of equations,
131
284330
2000
04:46
a set of equations that we can implement
132
286330
2000
04:48
on a chip. So it's just math.
133
288330
2000
04:50
In other words, we're not literally replacing
134
290330
3000
04:53
the components of the retina.
135
293330
2000
04:55
It's not like we're making a little mini-device
136
295330
2000
04:57
for each of the different cell types.
137
297330
2000
04:59
We've just abstracted what the
138
299330
2000
05:01
retina's doing with a set of equations.
139
301330
2000
05:03
And so, in a way, the equations are serving
140
303330
2000
05:05
as sort of a codebook. An image comes in,
141
305330
2000
05:07
goes through the set of equations,
142
307330
3000
05:10
and out comes streams of electrical pulses,
143
310330
2000
05:12
just like a normal retina would produce.
144
312330
4000
05:16
Now let me put my money
145
316330
2000
05:18
where my mouth is and show you that
146
318330
2000
05:20
we can actually produce normal output,
147
320330
2000
05:22
and what the implications of this are.
148
322330
2000
05:24
Here are three sets of
149
324330
2000
05:26
firing patterns. The top one is from
150
326330
2000
05:28
a normal animal, the middle one is from
151
328330
2000
05:30
a blind animal that's been treated with
152
330330
2000
05:32
this encoder-transducer device, and the
153
332330
2000
05:34
bottom one is from a blind animal treated
154
334330
2000
05:36
with a standard prosthetic.
155
336330
2000
05:38
So the bottom one is the state-of-the-art
156
338330
2000
05:40
device that's out there right now, which is
157
340330
2000
05:42
basically made up of light detectors,
158
342330
2000
05:44
but no encoder. So what we did was we
159
344330
2000
05:46
presented movies of everyday things --
160
346330
2000
05:48
people, babies, park benches,
161
348330
2000
05:50
you know, regular things happening -- and
162
350330
2000
05:52
we recorded the responses from the retinas
163
352330
2000
05:54
of these three groups of animals.
164
354330
2000
05:56
Now just to orient you, each box is showing
165
356330
2000
05:58
the firing patterns of several cells,
166
358330
2000
06:00
and just as in the previous slides,
167
360330
2000
06:02
each row is a different cell,
168
362330
2000
06:04
and I just made the pulses a little bit smaller
169
364330
2000
06:06
and thinner so I could show you
170
366330
3000
06:09
a long stretch of data.
171
369330
2000
06:11
So as you can see, the firing patterns
172
371330
2000
06:13
from the blind animal treated with
173
373330
2000
06:15
the encoder-transducer really do very
174
375330
2000
06:17
closely match the normal firing patterns --
175
377330
2000
06:19
and it's not perfect, but it's pretty good --
176
379330
2000
06:21
and the blind animal treated with
177
381330
2000
06:23
the standard prosthetic,
178
383330
2000
06:25
the responses really don't.
179
385330
2000
06:27
And so with the standard method,
180
387330
3000
06:30
the cells do fire, they just don't fire
181
390330
2000
06:32
in the normal firing patterns because
182
392330
2000
06:34
they don't have the right code.
183
394330
2000
06:36
How important is this?
184
396330
2000
06:38
What's the potential impact
185
398330
2000
06:40
on a patient's ability to see?
186
400330
3000
06:43
So I'm just going to show you one
187
403330
2000
06:45
bottom-line experiment that answers this,
188
405330
2000
06:47
and of course I've got a lot of other data,
189
407330
2000
06:49
so if you're interested I'm happy
190
409330
2000
06:51
to show more. So the experiment
191
411330
2000
06:53
is called a reconstruction experiment.
192
413330
2000
06:55
So what we did is we took a moment
193
415330
2000
06:57
in time from these recordings and asked,
194
417330
3000
07:00
what was the retina seeing at that moment?
195
420330
2000
07:02
Can we reconstruct what the retina
196
422330
2000
07:04
was seeing from the responses
197
424330
2000
07:06
from the firing patterns?
198
426330
2000
07:08
So, when we did this for responses
199
428330
3000
07:11
from the standard method and from
200
431330
3000
07:14
our encoder and transducer.
201
434330
2000
07:16
So let me show you, and I'm going to
202
436330
2000
07:18
start with the standard method first.
203
438330
2000
07:20
So you can see that it's pretty limited,
204
440330
2000
07:22
and because the firing patterns aren't
205
442330
2000
07:24
in the right code, they're very limited in
206
444330
2000
07:26
what they can tell you about
207
446330
2000
07:28
what's out there. So you can see that
208
448330
2000
07:30
there's something there, but it's not so clear
209
450330
2000
07:32
what that something is, and this just sort of
210
452330
2000
07:34
circles back to what I was saying in the
211
454330
2000
07:36
beginning, that with the standard method,
212
456330
2000
07:38
patients can see high-contrast edges, they
213
458330
2000
07:40
can see light, but it doesn't easily go
214
460330
2000
07:42
further than that. So what was
215
462330
2000
07:44
the image? It was a baby's face.
216
464330
3000
07:47
So what about with our approach,
217
467330
2000
07:49
adding the code? And you can see
218
469330
2000
07:51
that it's much better. Not only can you
219
471330
2000
07:53
tell that it's a baby's face, but you can
220
473330
2000
07:55
tell that it's this baby's face, which is a
221
475330
2000
07:57
really challenging task.
222
477330
2000
07:59
So on the left is the encoder
223
479330
2000
08:01
alone, and on the right is from an actual
224
481330
2000
08:03
blind retina, so the encoder and the transducer.
225
483330
2000
08:05
But the key one really is the encoder alone,
226
485330
2000
08:07
because we can team up the encoder with
227
487330
2000
08:09
the different transducer.
228
489330
2000
08:11
This is just actually the first one that we tried.
229
491330
2000
08:13
I just wanted to say something about the standard method.
230
493330
2000
08:15
When this first came out, it was just a really
231
495330
2000
08:17
exciting thing, the idea that you
232
497330
2000
08:19
even make a blind retina respond at all.
233
499330
3000
08:22
But there was this limiting factor,
234
502330
3000
08:25
the issue of the code, and how to make
235
505330
2000
08:27
the cells respond better,
236
507330
2000
08:29
produce normal responses,
237
509330
2000
08:31
and so this was our contribution.
238
511330
2000
08:33
Now I just want to wrap up,
239
513330
2000
08:35
and as I was mentioning earlier
240
515330
2000
08:37
of course I have a lot of other data
241
517330
2000
08:39
if you're interested, but I just wanted to give
242
519330
2000
08:41
this sort of basic idea
243
521330
2000
08:43
of being able to communicate
244
523330
3000
08:46
with the brain in its language, and
245
526330
2000
08:48
the potential power of being able to do that.
246
528330
3000
08:51
So it's different from the motor prosthetics
247
531330
2000
08:53
where you're communicating from the brain
248
533330
2000
08:55
to a device. Here we have to communicate
249
535330
2000
08:57
from the outside world
250
537330
2000
08:59
into the brain and be understood,
251
539330
2000
09:01
and be understood by the brain.
252
541330
2000
09:03
And then the last thing I wanted
253
543330
2000
09:05
to say, really, is to emphasize
254
545330
2000
09:07
that the idea generalizes.
255
547330
2000
09:09
So the same strategy that we used
256
549330
2000
09:11
to find the code for the retina we can also
257
551330
2000
09:13
use to find the code for other areas,
258
553330
2000
09:15
for example, the auditory system and
259
555330
2000
09:17
the motor system, so for treating deafness
260
557330
2000
09:19
and for motor disorders.
261
559330
2000
09:21
So just the same way that we were able to
262
561330
2000
09:23
jump over the damaged
263
563330
2000
09:25
circuitry in the retina to get to the retina's
264
565330
2000
09:27
output cells, we can jump over the
265
567330
2000
09:29
damaged circuitry in the cochlea
266
569330
2000
09:31
to get the auditory nerve,
267
571330
2000
09:33
or jump over damaged areas in the cortex,
268
573330
2000
09:35
in the motor cortex, to bridge the gap
269
575330
3000
09:38
produced by a stroke.
270
578330
2000
09:40
I just want to end with a simple
271
580330
2000
09:42
message that understanding the code
272
582330
2000
09:44
is really, really important, and if we
273
584330
2000
09:46
can understand the code,
274
586330
2000
09:48
the language of the brain, things become
275
588330
2000
09:50
possible that didn't seem obviously
276
590330
2000
09:52
possible before. Thank you.
277
592330
2000
09:54
(Applause)
278
594330
5000
About this website

This site will introduce you to YouTube videos that are useful for learning English. You will see English lessons taught by top-notch teachers from around the world. Double-click on the English subtitles displayed on each video page to play the video from there. The subtitles scroll in sync with the video playback. If you have any comments or requests, please contact us using this contact form.

https://forms.gle/WvT1wiN1qDtmnspy7