Anders Ynnerman: Visualizing the medical data explosion

42,202 views ・ 2011-01-21

TED


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

00:15
I will start by posing a little bit of a challenge:
0
15260
4000
00:19
the challenge of dealing with data,
1
19260
3000
00:22
data that we have to deal with
2
22260
2000
00:24
in medical situations.
3
24260
2000
00:26
It's really a huge challenge for us.
4
26260
2000
00:28
And this is our beast of burden --
5
28260
2000
00:30
this is a Computer Tomography machine,
6
30260
2000
00:32
a CT machine.
7
32260
2000
00:34
It's a fantastic device.
8
34260
2000
00:36
It uses X-rays, X-ray beams,
9
36260
2000
00:38
that are rotating very fast around the human body.
10
38260
3000
00:41
It takes about 30 seconds to go through the whole machine
11
41260
2000
00:43
and is generating enormous amounts of information
12
43260
2000
00:45
that comes out of the machine.
13
45260
2000
00:47
So this is a fantastic machine
14
47260
2000
00:49
that we can use
15
49260
2000
00:51
for improving health care,
16
51260
2000
00:53
but as I said, it's also a challenge for us.
17
53260
2000
00:55
And the challenge is really found in this picture here.
18
55260
3000
00:58
It's the medical data explosion
19
58260
2000
01:00
that we're having right now.
20
60260
2000
01:02
We're facing this problem.
21
62260
2000
01:04
And let me step back in time.
22
64260
2000
01:06
Let's go back a few years in time and see what happened back then.
23
66260
3000
01:09
These machines that came out --
24
69260
2000
01:11
they started coming in the 1970s --
25
71260
2000
01:13
they would scan human bodies,
26
73260
2000
01:15
and they would generate about 100 images
27
75260
2000
01:17
of the human body.
28
77260
2000
01:19
And I've taken the liberty, just for clarity,
29
79260
2000
01:21
to translate that to data slices.
30
81260
3000
01:24
That would correspond to about 50 megabytes of data,
31
84260
2000
01:26
which is small
32
86260
2000
01:28
when you think about the data we can handle today
33
88260
3000
01:31
just on normal mobile devices.
34
91260
2000
01:33
If you translate that to phone books,
35
93260
2000
01:35
it's about one meter of phone books in the pile.
36
95260
3000
01:38
Looking at what we're doing today
37
98260
2000
01:40
with these machines that we have,
38
100260
2000
01:42
we can, just in a few seconds,
39
102260
2000
01:44
get 24,000 images out of a body,
40
104260
2000
01:46
and that would correspond to about 20 gigabytes of data,
41
106260
3000
01:49
or 800 phone books,
42
109260
2000
01:51
and the pile would then be 200 meters of phone books.
43
111260
2000
01:53
What's about to happen --
44
113260
2000
01:55
and we're seeing this; it's beginning --
45
115260
2000
01:57
a technology trend that's happening right now
46
117260
2000
01:59
is that we're starting to look at time-resolved situations as well.
47
119260
3000
02:02
So we're getting the dynamics out of the body as well.
48
122260
3000
02:05
And just assume
49
125260
2000
02:07
that we will be collecting data during five seconds,
50
127260
3000
02:10
and that would correspond to one terabyte of data --
51
130260
2000
02:12
that's 800,000 books
52
132260
2000
02:14
and 16 kilometers of phone books.
53
134260
2000
02:16
That's one patient, one data set.
54
136260
2000
02:18
And this is what we have to deal with.
55
138260
2000
02:20
So this is really the enormous challenge that we have.
56
140260
3000
02:23
And already today -- this is 25,000 images.
57
143260
3000
02:26
Imagine the days
58
146260
2000
02:28
when we had radiologists doing this.
59
148260
2000
02:30
They would put up 25,000 images,
60
150260
2000
02:32
they would go like this, "25,0000, okay, okay.
61
152260
3000
02:35
There is the problem."
62
155260
2000
02:37
They can't do that anymore. That's impossible.
63
157260
2000
02:39
So we have to do something that's a little bit more intelligent than doing this.
64
159260
3000
02:43
So what we do is that we put all these slices together.
65
163260
2000
02:45
Imagine that you slice your body in all these directions,
66
165260
3000
02:48
and then you try to put the slices back together again
67
168260
3000
02:51
into a pile of data, into a block of data.
68
171260
2000
02:53
So this is really what we're doing.
69
173260
2000
02:55
So this gigabyte or terabyte of data, we're putting it into this block.
70
175260
3000
02:58
But of course, the block of data
71
178260
2000
03:00
just contains the amount of X-ray
72
180260
2000
03:02
that's been absorbed in each point in the human body.
73
182260
2000
03:04
So what we need to do is to figure out a way
74
184260
2000
03:06
of looking at the things we do want to look at
75
186260
3000
03:09
and make things transparent that we don't want to look at.
76
189260
3000
03:12
So transforming the data set
77
192260
2000
03:14
into something that looks like this.
78
194260
2000
03:16
And this is a challenge.
79
196260
2000
03:18
This is a huge challenge for us to do that.
80
198260
3000
03:21
Using computers, even though they're getting faster and better all the time,
81
201260
3000
03:24
it's a challenge to deal with gigabytes of data,
82
204260
2000
03:26
terabytes of data
83
206260
2000
03:28
and extracting the relevant information.
84
208260
2000
03:30
I want to look at the heart.
85
210260
2000
03:32
I want to look at the blood vessels. I want to look at the liver.
86
212260
2000
03:34
Maybe even find a tumor,
87
214260
2000
03:36
in some cases.
88
216260
2000
03:39
So this is where this little dear comes into play.
89
219260
2000
03:41
This is my daughter.
90
221260
2000
03:43
This is as of 9 a.m. this morning.
91
223260
2000
03:45
She's playing a computer game.
92
225260
2000
03:47
She's only two years old,
93
227260
2000
03:49
and she's having a blast.
94
229260
2000
03:51
So she's really the driving force
95
231260
3000
03:54
behind the development of graphics-processing units.
96
234260
3000
03:58
As long as kids are playing computer games,
97
238260
2000
04:00
graphics is getting better and better and better.
98
240260
2000
04:02
So please go back home, tell your kids to play more games,
99
242260
2000
04:04
because that's what I need.
100
244260
2000
04:06
So what's inside of this machine
101
246260
2000
04:08
is what enables me to do the things that I'm doing
102
248260
2000
04:10
with the medical data.
103
250260
2000
04:12
So really what I'm doing is using these fantastic little devices.
104
252260
3000
04:15
And you know, going back
105
255260
2000
04:17
maybe 10 years in time
106
257260
2000
04:19
when I got the funding
107
259260
2000
04:21
to buy my first graphics computer --
108
261260
2000
04:23
it was a huge machine.
109
263260
2000
04:25
It was cabinets of processors and storage and everything.
110
265260
3000
04:28
I paid about one million dollars for that machine.
111
268260
3000
04:32
That machine is, today, about as fast as my iPhone.
112
272260
3000
04:37
So every month there are new graphics cards coming out,
113
277260
2000
04:39
and here is a few of the latest ones from the vendors --
114
279260
3000
04:42
NVIDIA, ATI, Intel is out there as well.
115
282260
3000
04:45
And you know, for a few hundred bucks
116
285260
2000
04:47
you can get these things and put them into your computer,
117
287260
2000
04:49
and you can do fantastic things with these graphics cards.
118
289260
3000
04:52
So this is really what's enabling us
119
292260
2000
04:54
to deal with the explosion of data in medicine,
120
294260
3000
04:57
together with some really nifty work
121
297260
2000
04:59
in terms of algorithms --
122
299260
2000
05:01
compressing data,
123
301260
2000
05:03
extracting the relevant information that people are doing research on.
124
303260
3000
05:06
So I'm going to show you a few examples of what we can do.
125
306260
3000
05:09
This is a data set that was captured using a CT scanner.
126
309260
3000
05:12
You can see that this is a full data [set].
127
312260
3000
05:15
It's a woman. You can see the hair.
128
315260
3000
05:18
You can see the individual structures of the woman.
129
318260
3000
05:21
You can see that there is [a] scattering of X-rays
130
321260
3000
05:24
on the teeth, the metal in the teeth.
131
324260
2000
05:26
That's where those artifacts are coming from.
132
326260
3000
05:29
But fully interactively
133
329260
2000
05:31
on standard graphics cards on a normal computer,
134
331260
3000
05:34
I can just put in a clip plane.
135
334260
2000
05:36
And of course all the data is inside,
136
336260
2000
05:38
so I can start rotating, I can look at it from different angles,
137
338260
3000
05:41
and I can see that this woman had a problem.
138
341260
3000
05:44
She had a bleeding up in the brain,
139
344260
2000
05:46
and that's been fixed with a little stent,
140
346260
2000
05:48
a metal clamp that's tightening up the vessel.
141
348260
2000
05:50
And just by changing the functions,
142
350260
2000
05:52
then I can decide what's going to be transparent
143
352260
3000
05:55
and what's going to be visible.
144
355260
2000
05:57
I can look at the skull structure,
145
357260
2000
05:59
and I can see that, okay, this is where they opened up the skull on this woman,
146
359260
3000
06:02
and that's where they went in.
147
362260
2000
06:04
So these are fantastic images.
148
364260
2000
06:06
They're really high resolution,
149
366260
2000
06:08
and they're really showing us what we can do
150
368260
2000
06:10
with standard graphics cards today.
151
370260
3000
06:13
Now we have really made use of this,
152
373260
2000
06:15
and we have tried to squeeze a lot of data
153
375260
3000
06:18
into the system.
154
378260
2000
06:20
And one of the applications that we've been working on --
155
380260
2000
06:22
and this has gotten a little bit of traction worldwide --
156
382260
3000
06:25
is the application of virtual autopsies.
157
385260
2000
06:27
So again, looking at very, very large data sets,
158
387260
2000
06:29
and you saw those full-body scans that we can do.
159
389260
3000
06:32
We're just pushing the body through the whole CT scanner,
160
392260
3000
06:35
and just in a few seconds we can get a full-body data set.
161
395260
3000
06:38
So this is from a virtual autopsy.
162
398260
2000
06:40
And you can see how I'm gradually peeling off.
163
400260
2000
06:42
First you saw the body bag that the body came in,
164
402260
3000
06:45
then I'm peeling off the skin -- you can see the muscles --
165
405260
3000
06:48
and eventually you can see the bone structure of this woman.
166
408260
3000
06:51
Now at this point, I would also like to emphasize
167
411260
3000
06:54
that, with the greatest respect
168
414260
2000
06:56
for the people that I'm now going to show --
169
416260
2000
06:58
I'm going to show you a few cases of virtual autopsies --
170
418260
2000
07:00
so it's with great respect for the people
171
420260
2000
07:02
that have died under violent circumstances
172
422260
2000
07:04
that I'm showing these pictures to you.
173
424260
3000
07:08
In the forensic case --
174
428260
2000
07:10
and this is something
175
430260
2000
07:12
that ... there's been approximately 400 cases so far
176
432260
2000
07:14
just in the part of Sweden that I come from
177
434260
2000
07:16
that has been undergoing virtual autopsies
178
436260
2000
07:18
in the past four years.
179
438260
2000
07:20
So this will be the typical workflow situation.
180
440260
3000
07:23
The police will decide --
181
443260
2000
07:25
in the evening, when there's a case coming in --
182
445260
2000
07:27
they will decide, okay, is this a case where we need to do an autopsy?
183
447260
3000
07:30
So in the morning, in between six and seven in the morning,
184
450260
3000
07:33
the body is then transported inside of the body bag
185
453260
2000
07:35
to our center
186
455260
2000
07:37
and is being scanned through one of the CT scanners.
187
457260
2000
07:39
And then the radiologist, together with the pathologist
188
459260
2000
07:41
and sometimes the forensic scientist,
189
461260
2000
07:43
looks at the data that's coming out,
190
463260
2000
07:45
and they have a joint session.
191
465260
2000
07:47
And then they decide what to do in the real physical autopsy after that.
192
467260
3000
07:52
Now looking at a few cases,
193
472260
2000
07:54
here's one of the first cases that we had.
194
474260
2000
07:56
You can really see the details of the data set.
195
476260
3000
07:59
It's very high-resolution,
196
479260
2000
08:01
and it's our algorithms that allow us
197
481260
2000
08:03
to zoom in on all the details.
198
483260
2000
08:05
And again, it's fully interactive,
199
485260
2000
08:07
so you can rotate and you can look at things in real time
200
487260
2000
08:09
on these systems here.
201
489260
2000
08:11
Without saying too much about this case,
202
491260
2000
08:13
this is a traffic accident,
203
493260
2000
08:15
a drunk driver hit a woman.
204
495260
2000
08:17
And it's very, very easy to see the damages on the bone structure.
205
497260
3000
08:20
And the cause of death is the broken neck.
206
500260
3000
08:23
And this women also ended up under the car,
207
503260
2000
08:25
so she's quite badly beaten up
208
505260
2000
08:27
by this injury.
209
507260
2000
08:29
Here's another case, a knifing.
210
509260
3000
08:32
And this is also again showing us what we can do.
211
512260
2000
08:34
It's very easy to look at metal artifacts
212
514260
2000
08:36
that we can show inside of the body.
213
516260
3000
08:39
You can also see some of the artifacts from the teeth --
214
519260
3000
08:42
that's actually the filling of the teeth --
215
522260
2000
08:44
but because I've set the functions to show me metal
216
524260
3000
08:47
and make everything else transparent.
217
527260
2000
08:49
Here's another violent case. This really didn't kill the person.
218
529260
3000
08:52
The person was killed by stabs in the heart,
219
532260
2000
08:54
but they just deposited the knife
220
534260
2000
08:56
by putting it through one of the eyeballs.
221
536260
2000
08:58
Here's another case.
222
538260
2000
09:00
It's very interesting for us
223
540260
2000
09:02
to be able to look at things like knife stabbings.
224
542260
2000
09:04
Here you can see that knife went through the heart.
225
544260
3000
09:07
It's very easy to see how air has been leaking
226
547260
2000
09:09
from one part to another part,
227
549260
2000
09:11
which is difficult to do in a normal, standard, physical autopsy.
228
551260
3000
09:14
So it really, really helps
229
554260
2000
09:16
the criminal investigation
230
556260
2000
09:18
to establish the cause of death,
231
558260
2000
09:20
and in some cases also directing the investigation in the right direction
232
560260
3000
09:23
to find out who the killer really was.
233
563260
2000
09:25
Here's another case that I think is interesting.
234
565260
2000
09:27
Here you can see a bullet
235
567260
2000
09:29
that has lodged just next to the spine on this person.
236
569260
3000
09:32
And what we've done is that we've turned the bullet into a light source,
237
572260
3000
09:35
so that bullet is actually shining,
238
575260
2000
09:37
and it makes it really easy to find these fragments.
239
577260
3000
09:40
During a physical autopsy,
240
580260
2000
09:42
if you actually have to dig through the body to find these fragments,
241
582260
2000
09:44
that's actually quite hard to do.
242
584260
2000
09:48
One of the things that I'm really, really happy
243
588260
2000
09:50
to be able to show you here today
244
590260
3000
09:53
is our virtual autopsy table.
245
593260
2000
09:55
It's a touch device that we have developed
246
595260
2000
09:57
based on these algorithms, using standard graphics GPUs.
247
597260
3000
10:00
It actually looks like this,
248
600260
2000
10:02
just to give you a feeling for what it looks like.
249
602260
3000
10:05
It really just works like a huge iPhone.
250
605260
3000
10:08
So we've implemented
251
608260
2000
10:10
all the gestures you can do on the table,
252
610260
3000
10:13
and you can think of it as an enormous touch interface.
253
613260
4000
10:17
So if you were thinking of buying an iPad,
254
617260
2000
10:19
forget about it. This is what you want instead.
255
619260
3000
10:22
Steve, I hope you're listening to this, all right.
256
622260
3000
10:26
So it's a very nice little device.
257
626260
2000
10:28
So if you have the opportunity, please try it out.
258
628260
2000
10:30
It's really a hands-on experience.
259
630260
3000
10:33
So it gained some traction, and we're trying to roll this out
260
633260
3000
10:36
and trying to use it for educational purposes,
261
636260
2000
10:38
but also, perhaps in the future,
262
638260
2000
10:40
in a more clinical situation.
263
640260
3000
10:43
There's a YouTube video that you can download and look at this,
264
643260
2000
10:45
if you want to convey the information to other people
265
645260
2000
10:47
about virtual autopsies.
266
647260
3000
10:50
Okay, now that we're talking about touch,
267
650260
2000
10:52
let me move on to really "touching" data.
268
652260
2000
10:54
And this is a bit of science fiction now,
269
654260
2000
10:56
so we're moving into really the future.
270
656260
3000
10:59
This is not really what the medical doctors are using right now,
271
659260
3000
11:02
but I hope they will in the future.
272
662260
2000
11:04
So what you're seeing on the left is a touch device.
273
664260
3000
11:07
It's a little mechanical pen
274
667260
2000
11:09
that has very, very fast step motors inside of the pen.
275
669260
3000
11:12
And so I can generate a force feedback.
276
672260
2000
11:14
So when I virtually touch data,
277
674260
2000
11:16
it will generate forces in the pen, so I get a feedback.
278
676260
3000
11:19
So in this particular situation,
279
679260
2000
11:21
it's a scan of a living person.
280
681260
2000
11:23
I have this pen, and I look at the data,
281
683260
3000
11:26
and I move the pen towards the head,
282
686260
2000
11:28
and all of a sudden I feel resistance.
283
688260
2000
11:30
So I can feel the skin.
284
690260
2000
11:32
If I push a little bit harder, I'll go through the skin,
285
692260
2000
11:34
and I can feel the bone structure inside.
286
694260
3000
11:37
If I push even harder, I'll go through the bone structure,
287
697260
2000
11:39
especially close to the ear where the bone is very soft.
288
699260
3000
11:42
And then I can feel the brain inside, and this will be the slushy like this.
289
702260
3000
11:45
So this is really nice.
290
705260
2000
11:47
And to take that even further, this is a heart.
291
707260
3000
11:50
And this is also due to these fantastic new scanners,
292
710260
3000
11:53
that just in 0.3 seconds,
293
713260
2000
11:55
I can scan the whole heart,
294
715260
2000
11:57
and I can do that with time resolution.
295
717260
2000
11:59
So just looking at this heart,
296
719260
2000
12:01
I can play back a video here.
297
721260
2000
12:03
And this is Karljohan, one of my graduate students
298
723260
2000
12:05
who's been working on this project.
299
725260
2000
12:07
And he's sitting there in front of the Haptic device, the force feedback system,
300
727260
3000
12:10
and he's moving his pen towards the heart,
301
730260
3000
12:13
and the heart is now beating in front of him,
302
733260
2000
12:15
so he can see how the heart is beating.
303
735260
2000
12:17
He's taken the pen, and he's moving it towards the heart,
304
737260
2000
12:19
and he's putting it on the heart,
305
739260
2000
12:21
and then he feels the heartbeats from the real living patient.
306
741260
3000
12:24
Then he can examine how the heart is moving.
307
744260
2000
12:26
He can go inside, push inside of the heart,
308
746260
2000
12:28
and really feel how the valves are moving.
309
748260
3000
12:31
And this, I think, is really the future for heart surgeons.
310
751260
3000
12:34
I mean it's probably the wet dream for a heart surgeon
311
754260
3000
12:37
to be able to go inside of the patient's heart
312
757260
3000
12:40
before you actually do surgery,
313
760260
2000
12:42
and do that with high-quality resolution data.
314
762260
2000
12:44
So this is really neat.
315
764260
2000
12:47
Now we're going even further into science fiction.
316
767260
3000
12:50
And we heard a little bit about functional MRI.
317
770260
3000
12:53
Now this is really an interesting project.
318
773260
3000
12:56
MRI is using magnetic fields
319
776260
2000
12:58
and radio frequencies
320
778260
2000
13:00
to scan the brain, or any part of the body.
321
780260
3000
13:03
So what we're really getting out of this
322
783260
2000
13:05
is information of the structure of the brain,
323
785260
2000
13:07
but we can also measure the difference
324
787260
2000
13:09
in magnetic properties of blood that's oxygenated
325
789260
3000
13:12
and blood that's depleted of oxygen.
326
792260
3000
13:15
That means that it's possible
327
795260
2000
13:17
to map out the activity of the brain.
328
797260
2000
13:19
So this is something that we've been working on.
329
799260
2000
13:21
And you just saw Motts the research engineer, there,
330
801260
3000
13:24
going into the MRI system,
331
804260
2000
13:26
and he was wearing goggles.
332
806260
2000
13:28
So he could actually see things in the goggles.
333
808260
2000
13:30
So I could present things to him while he's in the scanner.
334
810260
3000
13:33
And this is a little bit freaky,
335
813260
2000
13:35
because what Motts is seeing is actually this.
336
815260
2000
13:37
He's seeing his own brain.
337
817260
3000
13:40
So Motts is doing something here,
338
820260
2000
13:42
and probably he is going like this with his right hand,
339
822260
2000
13:44
because the left side is activated
340
824260
2000
13:46
on the motor cortex.
341
826260
2000
13:48
And then he can see that at the same time.
342
828260
2000
13:50
These visualizations are brand new.
343
830260
2000
13:52
And this is something that we've been researching for a little while.
344
832260
3000
13:55
This is another sequence of Motts' brain.
345
835260
3000
13:58
And here we asked Motts to calculate backwards from 100.
346
838260
3000
14:01
So he's going "100, 97, 94."
347
841260
2000
14:03
And then he's going backwards.
348
843260
2000
14:05
And you can see how the little math processor is working up here in his brain
349
845260
3000
14:08
and is lighting up the whole brain.
350
848260
2000
14:10
Well this is fantastic. We can do this in real time.
351
850260
2000
14:12
We can investigate things. We can tell him to do things.
352
852260
2000
14:14
You can also see that his visual cortex
353
854260
2000
14:16
is activated in the back of the head,
354
856260
2000
14:18
because that's where he's seeing, he's seeing his own brain.
355
858260
2000
14:20
And he's also hearing our instructions
356
860260
2000
14:22
when we tell him to do things.
357
862260
2000
14:24
The signal is really deep inside of the brain as well,
358
864260
2000
14:26
and it's shining through,
359
866260
2000
14:28
because all of the data is inside this volume.
360
868260
2000
14:30
And in just a second here you will see --
361
870260
2000
14:32
okay, here. Motts, now move your left foot.
362
872260
2000
14:34
So he's going like this.
363
874260
2000
14:36
For 20 seconds he's going like that,
364
876260
2000
14:38
and all of a sudden it lights up up here.
365
878260
2000
14:40
So we've got motor cortex activation up there.
366
880260
2000
14:42
So this is really, really nice,
367
882260
2000
14:44
and I think this is a great tool.
368
884260
2000
14:46
And connecting also with the previous talk here,
369
886260
2000
14:48
this is something that we could use as a tool
370
888260
2000
14:50
to really understand
371
890260
2000
14:52
how the neurons are working, how the brain is working,
372
892260
2000
14:54
and we can do this with very, very high visual quality
373
894260
3000
14:57
and very fast resolution.
374
897260
3000
15:00
Now we're also having a bit of fun at the center.
375
900260
2000
15:02
So this is a CAT scan -- Computer Aided Tomography.
376
902260
3000
15:06
So this is a lion from the local zoo
377
906260
2000
15:08
outside of Norrkoping in Kolmarden, Elsa.
378
908260
3000
15:11
So she came to the center,
379
911260
2000
15:13
and they sedated her
380
913260
2000
15:15
and then put her straight into the scanner.
381
915260
2000
15:17
And then, of course, I get the whole data set from the lion.
382
917260
3000
15:20
And I can do very nice images like this.
383
920260
2000
15:22
I can peel off the layer of the lion.
384
922260
2000
15:24
I can look inside of it.
385
924260
2000
15:26
And we've been experimenting with this.
386
926260
2000
15:28
And I think this is a great application
387
928260
2000
15:30
for the future of this technology,
388
930260
2000
15:32
because there's very little known about the animal anatomy.
389
932260
3000
15:35
What's known out there for veterinarians is kind of basic information.
390
935260
3000
15:38
We can scan all sorts of things,
391
938260
2000
15:40
all sorts of animals.
392
940260
2000
15:42
The only problem is to fit it into the machine.
393
942260
3000
15:45
So here's a bear.
394
945260
2000
15:47
It was kind of hard to get it in.
395
947260
2000
15:49
And the bear is a cuddly, friendly animal.
396
949260
3000
15:52
And here it is. Here is the nose of the bear.
397
952260
3000
15:55
And you might want to cuddle this one,
398
955260
3000
15:58
until you change the functions and look at this.
399
958260
3000
16:01
So be aware of the bear.
400
961260
2000
16:03
So with that,
401
963260
2000
16:05
I'd like to thank all the people
402
965260
2000
16:07
who have helped me to generate these images.
403
967260
2000
16:09
It's a huge effort that goes into doing this,
404
969260
2000
16:11
gathering the data and developing the algorithms,
405
971260
3000
16:14
writing all the software.
406
974260
2000
16:16
So, some very talented people.
407
976260
3000
16:19
My motto is always, I only hire people that are smarter than I am
408
979260
3000
16:22
and most of these are smarter than I am.
409
982260
2000
16:24
So thank you very much.
410
984260
2000
16:26
(Applause)
411
986260
4000
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