ABOUT THE SPEAKER
Rosalind Picard - Scientist, inventor, entrepreneur
MIT Professor Rosalind Picard invents technologies that help people better understand emotion and behaviors that impact human health.

Why you should listen

Rosalind Picard wrote Affective Computing, a book that outlines how to give machines the skills of emotional intelligence, credited for launching the growing field of affective computing. Named one of seven "Tech SuperHeros to Watch" by CNN, Picard cofounded two companies that have commercialized technologies invented by her team at MIT: Empatica, which created the first AI-based smartwatch cleared by FDA in Neurology, and Affectiva, which provides emotion AI software. Picard teaches and leads research in affective computing at the MIT Media Lab in Cambridge; works as MIT's faculty chair for MindHandHeart, a campus-wide wellbeing initiative; and serves as chief scientist for Empatica, creating AI-based analytics and wearables to improve human wellbeing.

More profile about the speaker
Rosalind Picard | Speaker | TED.com
TEDxBeaconStreet

Rosalind Picard: An AI smartwatch that detects seizures

Filmed:
1,872,363 views

Every year worldwide, more than 50,000 otherwise healthy people with epilepsy suddenly die -- a condition known as SUDEP. These deaths may be largely preventable, says AI researcher Rosalind Picard. Learn how Picard helped develop a cutting-edge smartwatch that can detect epileptic seizures before they occur and alert nearby loved ones in time to help.
- Scientist, inventor, entrepreneur
MIT Professor Rosalind Picard invents technologies that help people better understand emotion and behaviors that impact human health. Full bio

Double-click the English transcript below to play the video.

00:13
This is Henry,
0
1747
1487
00:15
a cute boy,
1
3258
1806
00:17
and when Henry was three,
2
5088
2138
00:19
his mom found him having
some febrile seizures.
3
7250
4935
00:25
Febrile seizures are seizures that occur
when you also have a fever,
4
13080
4960
00:30
and the doctor said,
5
18064
1898
00:31
"Don't worry too much.
Kids usually outgrow these."
6
19986
2742
00:35
When he was four,
he had a convulsive seizure,
7
23506
3981
00:39
the kind that you lose
consciousness and shake --
8
27511
3262
00:42
a generalized tonic-clonic seizure --
9
30797
2896
00:45
and while the diagnosis of epilepsy
was in the mail,
10
33717
6984
00:53
Henry's mom went to get him
out of bed one morning,
11
41763
2676
00:57
and as she went in his room,
12
45324
1904
01:00
she found his cold, lifeless body.
13
48753
4690
01:07
Henry died of SUDEP,
14
55904
2056
01:09
sudden unexpected death in epilepsy.
15
57984
2405
01:13
I'm curious how many of you
have heard of SUDEP.
16
61714
3245
01:17
This is a very well-educated audience,
and I see only a few hands.
17
65603
4366
01:21
SUDEP is when an otherwise
healthy person with epilepsy
18
69993
3966
01:25
dies and they can't attribute it
to anything they can find in an autopsy.
19
73983
3890
01:32
There is a SUDEP
every seven to nine minutes.
20
80082
3603
01:36
That's on average two per TED Talk.
21
84217
2879
01:43
Now, a normal brain
has electrical activity.
22
91411
3659
01:47
You can see some of the electrical waves
23
95094
2135
01:49
coming out of this picture
of a brain here.
24
97253
3064
01:52
And these should look
like typical electrical activity
25
100341
3746
01:56
that an EEG could read on the surface.
26
104111
2135
01:58
When you have a seizure,
it's a bit of unusual electrical activity,
27
106270
4417
02:02
and it can be focal.
28
110711
1385
02:04
It can take place
in just a small part of your brain.
29
112120
2500
02:06
When that happens,
you might have a strange sensation.
30
114644
3808
02:10
Several could be happening
here in the audience right now,
31
118476
3033
02:13
and the person next to you
might not even know.
32
121533
2602
02:16
However, if you have a seizure
where that little brush fire spreads
33
124159
3920
02:20
like a forest fire over the brain,
34
128103
1992
02:22
then it generalizes,
35
130119
1674
02:23
and that generalized seizure
takes your consciousness away
36
131817
4571
02:28
and causes you to convulse.
37
136412
1576
02:30
There are more SUDEPs
in the United States every year
38
138680
4142
02:34
than sudden infant death syndrome.
39
142846
2659
02:38
Now, how many of you have heard
of sudden infant death syndrome?
40
146298
3002
02:41
Right? Pretty much every hand goes up.
41
149324
1956
02:43
So what's going on here?
42
151304
1723
02:45
Why is this so much more common
and yet people haven't heard of it?
43
153051
5143
02:50
And what can you do to prevent it?
44
158218
2436
02:52
Well, there are two things,
scientifically shown,
45
160678
2302
02:55
that prevent or reduce the risk of SUDEP.
46
163004
3373
02:59
The first is: "Follow
your doctor's instructions,
47
167115
2555
03:01
take your medications."
48
169694
1436
03:03
Two-thirds of people who have epilepsy
49
171154
2201
03:05
get it under control
with their medications.
50
173379
2119
03:07
The second thing that reduces
the risk of SUDEP is companionship.
51
175848
4171
03:12
It's having somebody there
at the time that you have a seizure.
52
180043
4694
03:16
Now, SUDEP, even though
most of you have never heard of it,
53
184761
3921
03:20
is actually the number two cause
of years of potential life lost
54
188706
5189
03:25
of all neurological disorders.
55
193919
2976
03:29
The vertical axis is the number of deaths
56
197197
4286
03:33
times the remaining life span,
57
201507
3166
03:36
so higher is much worse impact.
58
204697
3439
03:40
SUDEP, however, unlike these others,
59
208522
2974
03:43
is something that people right here
could do something to push that down.
60
211520
5716
03:49
Now, what is Roz Picard, an AI researcher,
doing here telling you about SUDEP, right?
61
217778
7000
03:56
I'm not a neurologist.
62
224802
2521
03:59
When I was working at the Media Lab
on measurement of emotion,
63
227347
4215
04:03
trying to make our machines
more intelligent about our emotions,
64
231586
3222
04:06
we started doing a lot of work
measuring stress.
65
234832
2531
04:11
We built lots of sensors
66
239212
1936
04:13
that measured it
in lots of different ways.
67
241172
2381
04:15
But one of them in particular
68
243577
2320
04:17
grew out of some of this very old work
with measuring sweaty palms
69
245921
4112
04:22
with an electrical signal.
70
250057
1744
04:23
This is a signal of skin conductance
71
251825
1749
04:25
that's known to go up
when you get nervous,
72
253598
2118
04:27
but it turns out it also goes up with
a lot of other interesting conditions.
73
255740
3572
04:31
But measuring it with wires on your hand
is really inconvenient.
74
259336
3022
04:34
So we invented a bunch of other ways
of doing this at the MIT Media Lab.
75
262382
3745
04:38
And with these wearables,
76
266151
2173
04:40
we started to collect the first-ever
clinical quality data 24-7.
77
268348
4805
04:45
Here's a picture of what that looked like
78
273609
2533
04:48
the first time an MIT student collected
skin conductance on the wrist 24-7.
79
276166
6595
04:55
Let's zoom in a little bit here.
80
283187
2358
04:57
What you see is 24 hours
from left to right,
81
285569
3025
05:00
and here is two days of data.
82
288618
1927
05:02
And first, what surprised us
83
290569
2682
05:05
was sleep was the biggest
peak of the day.
84
293275
3346
05:08
Now, that sounds broken, right?
85
296645
1750
05:10
You're calm when you're asleep,
so what's going on here?
86
298419
4178
05:14
Well, it turns out
that our physiology during sleep
87
302621
2453
05:17
is very different
than our physiology during wake,
88
305098
2643
05:19
and while there's still a bit of a mystery
89
307765
2184
05:21
why these peaks are usually
the biggest of the day during sleep,
90
309973
3744
05:25
we now believe they're related
to memory consolidation
91
313741
3170
05:28
and memory formation during sleep.
92
316935
2055
05:31
We also saw things
that were exactly what we expected.
93
319895
3571
05:35
When an MIT student
is working hard in the lab
94
323490
2350
05:37
or on homeworks,
95
325864
1389
05:39
there is not only emotional stress,
but there's cognitive load,
96
327277
4547
05:43
and it turns out that cognitive load,
cognitive effort, mental engagement,
97
331848
4861
05:48
excitement about learning something --
98
336733
2194
05:50
those things also make the signal go up.
99
338951
2337
05:55
Unfortunately, to the embarrassment
of we MIT professors,
100
343014
3738
05:58
(Laughter)
101
346776
1079
05:59
the low point every day
is classroom activity.
102
347879
4468
06:04
Now, I am just showing you
one person's data here,
103
352810
2516
06:07
but this, unfortunately,
is true in general.
104
355350
2610
06:12
This sweatband has inside it
a homebuilt skin-conductance sensor,
105
360400
4389
06:16
and one day, one of our undergrads
knocked on my door
106
364813
5086
06:21
right at the end of the December semester,
107
369923
2675
06:24
and he said, "Professor Picard,
108
372622
2190
06:26
can I please borrow
one of your wristband sensors?
109
374836
2714
06:29
My little brother has autism,
he can't talk,
110
377574
3729
06:33
and I want to see
what's stressing him out."
111
381327
2167
06:36
And I said, "Sure, in fact,
don't just take one, take two,"
112
384434
2858
06:39
because they broke easily back then.
113
387316
2792
06:42
So he took them home,
he put them on his little brother.
114
390132
2620
06:44
Now, I was back in MIT,
looking at the data on my laptop,
115
392776
3169
06:47
and the first day, I thought,
"Hmm, that's odd,
116
395969
2861
06:50
he put them on both wrists
instead of waiting for one to break.
117
398854
2991
06:53
OK, fine, don't follow my instructions."
118
401869
2320
06:56
I'm glad he didn't.
119
404213
1718
06:58
Second day -- chill.
Looked like classroom activity.
120
406377
3206
07:01
(Laughter)
121
409607
1262
07:03
A few more days ahead.
122
411389
1329
07:05
The next day, one wrist signal was flat
123
413404
4514
07:09
and the other had
the biggest peak I've ever seen,
124
417942
3958
07:14
and I thought, "What's going on?
125
422496
2318
07:16
We've stressed people out at MIT
every way imaginable.
126
424838
3460
07:20
I've never seen a peak this big."
127
428933
1738
07:24
And it was only on one side.
128
432409
1969
07:26
How can you be stressed on one side
of your body and not the other?
129
434402
3144
07:29
So I thought one or both sensors
must be broken.
130
437570
2870
07:33
Now, I'm an electroengineer by training,
131
441185
1917
07:35
so I started a whole bunch of stuff
to try to debug this,
132
443126
2669
07:37
and long story short,
I could not reproduce this.
133
445819
2499
07:40
So I resorted to old-fashioned debugging.
134
448962
3033
07:44
I called the student at home on vacation.
135
452019
3282
07:47
"Hi, how's your little brother?
How's your Christmas?
136
455325
4343
07:51
Hey, do you have any idea
what happened to him?"
137
459692
3281
07:54
And I gave this particular date and time,
138
462997
1956
07:56
and the data.
139
464977
1301
07:58
And he said, "I don't know,
I'll check the diary."
140
466302
3415
08:02
Diary? An MIT student keeps a diary?
141
470850
2908
08:05
So I waited and he came back.
142
473782
1819
08:07
He had the exact date and time,
143
475625
1539
08:09
and he says, "That was right before
he had a grand mal seizure."
144
477188
3853
08:14
Now, at the time, I didn't know
anything about epilepsy,
145
482942
3349
08:18
and did a bunch of research,
146
486315
2159
08:20
realized that another student's dad
is chief of neurosurgery
147
488498
3126
08:23
at Children's Hospital Boston,
148
491648
2000
08:25
screwed up my courage
and called Dr. Joe Madsen.
149
493672
2289
08:27
"Hi, Dr. Madsen,
my name's Rosalind Picard.
150
495985
2469
08:30
Is it possible somebody could have
151
498478
3896
08:34
a huge sympathetic
nervous system surge" --
152
502398
4403
08:38
that's what drives the skin conductance --
153
506825
2578
08:41
"20 minutes before a seizure?"
154
509427
1672
08:44
And he says, "Probably not."
155
512172
2119
08:47
He says, "It's interesting.
156
515997
1334
08:49
We've had people whose hair
stands on end on one arm
157
517355
2768
08:52
20 minutes before a seizure."
158
520147
2080
08:55
And I'm like, "On one arm?"
159
523060
1315
08:56
I didn't want to tell him that, initially,
160
524399
2130
08:58
because I thought this was too ridiculous.
161
526553
2389
09:00
He explained how this could
happen in the brain,
162
528966
2293
09:03
and he got interested.
I showed him the data.
163
531283
2173
09:05
We made a whole bunch more devices,
got them safety certified.
164
533480
2948
09:08
90 families were being
enrolled in a study,
165
536452
2074
09:10
all with children who were going
to be monitored 24-7
166
538550
3229
09:13
with gold-standard EEG on their scalp
167
541803
3371
09:17
for reading the brain activity,
168
545198
1543
09:18
video to watch the behavior,
169
546765
1943
09:20
electrocardiogram -- ECG --
and now EDA, electrodermal activity,
170
548732
3660
09:24
to see if there was
something in this periphery
171
552416
2585
09:27
that we could easily pick up,
related to a seizure.
172
555025
2446
09:30
We found, in 100 percent
of the first batch of grand mal seizures,
173
558263
6441
09:36
this whopper of responses
in the skin conductance.
174
564728
3633
09:40
The blue in the middle, the boy's sleep,
175
568385
2003
09:42
is usually the biggest peak of the day.
176
570412
1972
09:44
These three seizures you see here
are popping out of the forest
177
572408
3632
09:48
like redwood trees.
178
576064
1476
09:51
Furthermore, when you couple
the skin conductance at the top
179
579096
3580
09:54
with the movement from the wrist
180
582700
3517
09:58
and you get lots of data
and train machine learning and AI on it,
181
586241
4800
10:03
you can build an automated AI
that detects these patterns
182
591065
4793
10:07
much better than just
a shake detector can do.
183
595882
4215
10:12
So we realized that we needed
to get this out,
184
600121
4267
10:16
and with the PhD work of Ming-Zher Poh
185
604412
2358
10:18
and later great improvements by Empatica,
186
606794
3027
10:21
this has made progress and the seizure
detection is much more accurate.
187
609845
3844
10:25
But we also learned some other things
about SUDEP during this.
188
613713
2905
10:28
One thing we learned is that SUDEP,
189
616642
3412
10:32
while it's rare after
a generalized tonic-clonic seizure,
190
620078
2912
10:35
that's when it's most likely
to happen -- after that type.
191
623014
3237
10:38
And when it happens,
it doesn't happen during the seizure,
192
626275
3119
10:41
and it doesn't usually happen
immediately afterwards,
193
629418
2866
10:44
but immediately afterwards,
194
632308
1715
10:46
when the person just seems
very still and quiet,
195
634047
3062
10:49
they may go into another phase,
where the breathing stops,
196
637133
4947
10:54
and then after the breathing stops,
later the heart stops.
197
642104
3024
10:57
So there's some time
to get somebody there.
198
645152
2043
11:00
We also learned that there is a region
deep in the brain called the amygdala,
199
648349
4904
11:05
which we had been studying
in our emotion research a lot.
200
653277
2936
11:08
We have two amygdalas,
201
656237
1281
11:09
and if you stimulate the right one,
202
657542
1673
11:11
you get a big right
skin conductance response.
203
659239
2739
11:14
Now, you have to sign up right now
for a craniotomy to get this done,
204
662002
4245
11:18
not exactly something
we're going to volunteer to do,
205
666271
2833
11:21
but it causes a big right skin
conductance response.
206
669128
2513
11:23
Stimulate the left one, big left
skin conductance response on the palm.
207
671665
3853
11:27
And furthermore, when somebody
stimulates your amygdala
208
675542
4040
11:31
while you're sitting there
and you might just be working,
209
679606
3419
11:35
you don't show any signs of distress,
210
683049
1888
11:37
but you stop breathing,
211
685620
1286
11:40
and you don't start again
until somebody stimulates you.
212
688470
3317
11:44
"Hey, Roz, are you there?"
213
692599
1513
11:46
And you open your mouth to talk.
214
694136
1931
11:48
As you take that breath to speak,
215
696868
2194
11:51
you start breathing again.
216
699086
1505
11:54
So we had started with work on stress,
217
702762
3620
11:58
which had enabled us
to build lots of sensors
218
706406
2253
12:00
that were gathering
high quality enough data
219
708683
2132
12:02
that we could leave the lab
and start to get this in the wild;
220
710839
2921
12:05
accidentally found a whopper
of a response with the seizure,
221
713784
2864
12:08
neurological activation that can cause
a much bigger response
222
716672
2859
12:11
than traditional stressors;
223
719555
1310
12:12
lots of partnership with hospitals
and an epilepsy monitoring unit,
224
720889
3198
12:16
especially Children's Hospital Boston
225
724111
1828
12:17
and the Brigham;
226
725963
1151
12:19
and machine learning and AI on top of this
227
727138
2525
12:21
to take and collect lots more data
228
729687
3021
12:24
in service of trying
to understand these events
229
732732
2629
12:27
and if we could prevent SUDEP.
230
735385
1594
12:30
This is now commercialized by Empatica,
231
738023
3563
12:33
a start-up that I had
the privilege to cofound,
232
741610
2889
12:36
and the team there has done an amazing job
improving the technology
233
744523
4206
12:40
to make a very beautiful sensor
234
748753
1759
12:42
that not only tells time and does steps
and sleep and all that good stuff,
235
750536
3859
12:46
but this is running real-time
AI and machine learning
236
754419
3096
12:49
to detect generalized
tonic-clonic seizures
237
757539
2618
12:52
and send an alert for help
238
760181
2151
12:54
if I were to have a seizure
and lose consciousness.
239
762356
3192
12:57
This just got FDA-approved
240
765572
2816
13:00
as the first smartwatch
to get approved in neurology.
241
768412
4140
13:05
(Applause)
242
773923
7000
13:14
Now, the next slide is what made
my skin conductance go up.
243
782674
3531
13:18
One morning, I'm checking my email
244
786811
1706
13:20
and I see a story from a mom
245
788541
2297
13:22
who said she was in the shower,
246
790862
1886
13:24
and her phone was
on the counter by the shower,
247
792772
2453
13:27
and it said her daughter
might need her help.
248
795249
2157
13:30
So she interrupts her shower and goes
running to her daughter's bedroom,
249
798196
3381
13:33
and she finds her daughter
facedown in bed, blue and not breathing.
250
801601
3743
13:37
She flips her over -- human stimulation --
251
805368
3281
13:40
and her daughter takes a breath,
and another breath,
252
808673
3683
13:44
and her daughter turns pink and is fine.
253
812380
3954
13:49
I think I turned white reading this email.
254
817551
3224
13:52
My first response is,
"Oh no, it's not perfect.
255
820799
2322
13:55
The Bluetooth could break,
the battery could die.
256
823145
2325
13:57
All these things could go wrong.
Don't rely on this."
257
825494
2477
13:59
And she said, "It's OK.
I know no technology is perfect.
258
827995
3408
14:03
None of us can always
be there all the time.
259
831427
2163
14:06
But this, this device plus AI
260
834844
4342
14:11
enabled me to get there in time
to save my daughter's life."
261
839210
3104
14:17
Now, I've been mentioning children,
262
845933
1906
14:19
but SUDEP peaks, actually,
among people in their 20s, 30s and 40s,
263
847863
5707
14:25
and the next line I'm going to put up
264
853594
1816
14:27
is probably going to make
some people uncomfortable,
265
855434
2453
14:29
but it's less uncomfortable
than we'll all be
266
857911
2374
14:32
if this list is extended
to somebody you know.
267
860309
3070
14:36
Could this happen to somebody you know?
268
864437
2338
14:39
And the reason I bring up
this uncomfortable question
269
867197
2742
14:41
is because one in 26 of you
will have epilepsy at some point,
270
869963
4912
14:46
and from what I've been learning,
271
874899
2235
14:49
people with epilepsy often don't tell
their friends and their neighbors
272
877158
3334
14:52
that they have it.
273
880516
1158
14:53
So if you're willing to let them
use an AI or whatever
274
881698
4921
14:58
to summon you in a moment
of possible need,
275
886643
4414
15:03
if you would let them know that,
276
891081
1618
15:04
you could make a difference in their life.
277
892723
2354
15:07
Why do all this hard work to build AIs?
278
895728
2851
15:12
A couple of reasons here:
279
900101
1201
15:13
one is Natasha, the girl who lived,
280
901326
2163
15:15
and her family wanted me
to tell you her name.
281
903513
2392
15:18
Another is her family
282
906977
1612
15:20
and the wonderful people out there
283
908613
2191
15:22
who want to be there to support people
who have conditions
284
910828
2825
15:25
that they've felt uncomfortable
in the past mentioning to others.
285
913677
3293
15:29
And the other reason is all of you,
286
917901
1873
15:31
because we have the opportunity
to shape the future of AI.
287
919798
5332
15:37
We can actually change it,
288
925154
2357
15:39
because we are the ones building it.
289
927535
2404
15:41
So let's build AI
290
929963
1977
15:43
that makes everybody's lives better.
291
931964
3292
15:47
Thank you.
292
935982
1151
15:49
(Applause)
293
937157
4414

▲Back to top

ABOUT THE SPEAKER
Rosalind Picard - Scientist, inventor, entrepreneur
MIT Professor Rosalind Picard invents technologies that help people better understand emotion and behaviors that impact human health.

Why you should listen

Rosalind Picard wrote Affective Computing, a book that outlines how to give machines the skills of emotional intelligence, credited for launching the growing field of affective computing. Named one of seven "Tech SuperHeros to Watch" by CNN, Picard cofounded two companies that have commercialized technologies invented by her team at MIT: Empatica, which created the first AI-based smartwatch cleared by FDA in Neurology, and Affectiva, which provides emotion AI software. Picard teaches and leads research in affective computing at the MIT Media Lab in Cambridge; works as MIT's faculty chair for MindHandHeart, a campus-wide wellbeing initiative; and serves as chief scientist for Empatica, creating AI-based analytics and wearables to improve human wellbeing.

More profile about the speaker
Rosalind Picard | Speaker | TED.com

Data provided by TED.

This site was created in May 2015 and the last update was on January 12, 2020. It will no longer be updated.

We are currently creating a new site called "eng.lish.video" and would be grateful if you could access it.

If you have any questions or suggestions, please feel free to write comments in your language on the contact form.

Privacy Policy

Developer's Blog

Buy Me A Coffee