An AI smartwatch that detects seizures | Rosalind Picard

翻译人员: Jingle duan 校对人员: Cissy Yun

This is Henry,

他叫亨利。

a cute boy,

一个很可爱的男孩。

and when Henry was three,

在他三岁的时候，

his mom found him having some febrile seizures.

他的母亲发现他患有发热性痉挛。

Febrile seizures are seizures that occur when you also have a fever,

发热性痉挛是在发热时 产生的癫痫症状。

and the doctor said,

医生说，

"Don't worry too much. Kids usually outgrow these."

“不用太过担心， 儿童一般在长大后症状就会消失。”

When he was four, he had a convulsive seizure,

四岁的时候，他出现了抽搐型癫痫。

the kind that you lose consciousness and shake --

就是那种让人失去意识， 并且颤抖不止的癫痫——

a generalized tonic-clonic seizure --

全身强直型阵挛性癫痫。

and while the diagnosis of epilepsy was in the mail,

当癫痫的诊断书还没寄到的时候，

Henry's mom went to get him out of bed one morning,

一天早晨，亨利的妈妈去叫他起床。

and as she went in his room,

进入到他的寝室时，

she found his cold, lifeless body.

她发现了他冰冷、无生命体征的身体。

Henry died of SUDEP,

亨利死于 SUDEP：

sudden unexpected death in epilepsy.

即癫痫猝死症。

I'm curious how many of you have heard of SUDEP.

我很好奇你们当中 有多少人听说过 SUDEP。

This is a very well-educated audience, and I see only a few hands.

在座各位都受过良好的教育， 而我却只看到了少数人举手。

SUDEP is when an otherwise healthy person with epilepsy

SUDEP 会让一个原本健康的 癫痫病人死亡后，

dies and they can't attribute it to anything they can find in an autopsy.

法医在其尸检中找不到 任何可以致死的原因。

There is a SUDEP every seven to nine minutes.

每7到9分钟就会有一例 SUDEP 发生。

That's on average two per TED Talk.

也就是说平均一场 TED 演讲的时间 会发生两起 SUDEP。

Now, a normal brain has electrical activity.

正常的大脑都会有电流产生。

You can see some of the electrical waves

大家可以在这张大脑的图片中，

coming out of this picture of a brain here.

看到一些脑电波。

And these should look like typical electrical activity

它们看起来和脑电图

that an EEG could read on the surface.

在头皮部位读取到的普通的电流一样。

When you have a seizure, it's a bit of unusual electrical activity,

一旦遇到癫痫发作， 你的脑电波会有异常，

and it can be focal.

这可能就是病灶所在。

It can take place in just a small part of your brain.

它可能只发生于 你大脑中的一小部分区域。

When that happens, you might have a strange sensation.

当它发生时， 你可能会有一种奇怪的感觉。

Several could be happening here in the audience right now,

就在当下的听众中， 现在可能就在发生着几例。

and the person next to you might not even know.

而你邻座的人并不知道。

However, if you have a seizure where that little brush fire spreads

然而，如果你的癫痫像小火苗触发的

like a forest fire over the brain,

森林大火一样在大脑中蔓延开，

then it generalizes,

这时，它就形成了一般意义上的癫痫。

and that generalized seizure takes your consciousness away

这种癫痫会带走你的意识

and causes you to convulse.

并且引起你的抽搐。

There are more SUDEPs in the United States every year

美国每年发生的 SUDEP 比

than sudden infant death syndrome.

婴儿突发综合症死亡数量还多。

Now, how many of you have heard of sudden infant death syndrome?

你们当中有多少人听说过 婴儿突发死亡综合症？

Right? Pretty much every hand goes up.

对吧? 几乎每个人都举起手了。

So what's going on here?

所以，这里到底发生了什么？

Why is this so much more common and yet people haven't heard of it?

为什么癫痫这么普遍， 人们却没有听说过它呢？

And what can you do to prevent it?

我们可以做什么来阻止癫痫呢？

Well, there are two things, scientifically shown,

好的，科学证明有两件事情

that prevent or reduce the risk of SUDEP.

可以阻止或者减少 SUDEP 的风险。

The first is: "Follow your doctor's instructions,

第一件事情就是： “遵循医嘱，

take your medications."

按时吃药。”

Two-thirds of people who have epilepsy

三分之二的癫痫病患者，

get it under control with their medications.

通过药物可以把病情控制住。

The second thing that reduces the risk of SUDEP is companionship.

第二件可以减少 SUDEP 风险的事情 是陪伴。

It's having somebody there at the time that you have a seizure.

也就是说在你发生癫痫时 有人陪在身边。

Now, SUDEP, even though most of you have never heard of it,

现在，虽然你们中的大多数人 从未听说过SUDEP，

is actually the number two cause of years of potential life lost

是 在神经系统疾病患者的 潜在死因排行榜上

of all neurological disorders.

SUDEP 多年位居第二。

The vertical axis is the number of deaths

竖轴是死亡人数，

times the remaining life span,

乘以剩余寿命。

so higher is much worse impact.

越高越糟糕。

SUDEP, however, unlike these others,

然而，SUDEP不像其它症状，

is something that people right here could do something to push that down.

在座各位可以做一些小事 就能把它降低。

Now, what is Roz Picard, an AI researcher, doing here telling you about SUDEP, right?

我，罗兹·皮卡德，一个AI研究者 能告诉你们有关 SUDEP 什么呢？

I'm not a neurologist.

我又不是一个神经学家。

When I was working at the Media Lab on measurement of emotion,

当我在研究情绪测量的 多媒体实验室工作时,

trying to make our machines more intelligent about our emotions,

我们研究如何让机器对 我们的情感更智能。

we started doing a lot of work measuring stress.

为此，我们启动了很多测量压力的工作。

We built lots of sensors

我们构建了很多传感器

that measured it in lots of different ways.

以很多不同的方式来测量压力。

But one of them in particular

但脱颖而出的是

grew out of some of this very old work with measuring sweaty palms

一个用电子信号来测量手掌出汗的

with an electrical signal.

老研究。

This is a signal of skin conductance

这是一个皮肤电导率的信号，

that's known to go up when you get nervous,

很多人知道， 它在你紧张时会上升，

but it turns out it also goes up with a lot of other interesting conditions.

它也会随着其它有趣的环境而上升。

But measuring it with wires on your hand is really inconvenient.

但是通过手上绑一根电线的方式 来测量确实很不方便。

So we invented a bunch of other ways of doing this at the MIT Media Lab.

所以我们发明了其它的方法 在MIT的多媒体实验室进行测试。

And with these wearables,

这些可穿戴测量仪让我们

we started to collect the first-ever clinical quality data 24-7.

能一周七天，24小时无间断地

获取这些临床质量数据， 这是前所未有的。

Here's a picture of what that looked like

这张照片就是

一个MIT在第一次7/24周期 从手腕上收集到的皮肤电导率。

the first time an MIT student collected skin conductance on the wrist 24-7.

Let's zoom in a little bit here.

我们放大这里。

What you see is 24 hours from left to right,

你们可以看到从左到右是24小时

and here is two days of data.

这里有两天的数据。

And first, what surprised us

首先，令我们震惊的是

was sleep was the biggest peak of the day.

睡眠时，数据达到一天中最高峰。

Now, that sounds broken, right?

现在，听上去好像是仪器坏了？

You're calm when you're asleep, so what's going on here?

你在睡觉时是平静的， 所以，究竟发生了什么？

Well, it turns out that our physiology during sleep

研究发现我们在睡眠时的生理机能

is very different than our physiology during wake,

与我们在清醒时的生理机能大为不同。

and while there's still a bit of a mystery

虽然我们对此的了解还不全面，

why these peaks are usually the biggest of the day during sleep,

“为什么通常总是在睡眠时 出现一天的最高峰？”

we now believe they're related to memory consolidation

我们现在认为它们和睡眠时的记忆固化

and memory formation during sleep.

以及记忆形成有关。

We also saw things that were exactly what we expected.

我们同样看到了我们精确期待 会发生的事情。

When an MIT student is working hard in the lab

当一个MIT的学生在努力在实验室工作

or on homeworks,

或者在写作业时，

there is not only emotional stress, but there's cognitive load,

他承受的不仅有情感压力， 还有认知负载，

and it turns out that cognitive load, cognitive effort, mental engagement,

研究表明认知负载，认知努力， 精神上的专注

excitement about learning something --

还有对所学习东西的兴奋等,

those things also make the signal go up.

这些东西都会让这个信号上升。

Unfortunately, to the embarrassment of we MIT professors,

很不幸， 这让我们MIT的教授们有些难堪，

(Laughter)

（大笑）

the low point every day is classroom activity.

因为，每天的低点是课堂活动。

Now, I am just showing you one person's data here,

我在这里只给你们展示了一个人的数据。

but this, unfortunately, is true in general.

但是这点，很不幸， 却基本上是真实的。

This sweatband has inside it a homebuilt skin-conductance sensor,

这个防汗带内置了一个自主研发的 皮肤电传导传感器，

and one day, one of our undergrads knocked on my door

一天，我们的一个大学生

right at the end of the December semester,

在12月学期结束的时候敲开了我的门

and he said, "Professor Picard,

然后他说到：“皮卡德教授

can I please borrow one of your wristband sensors?

我能借一个您的腕部传感器吗？

My little brother has autism, he can't talk,

我幼小的弟弟有自闭症， 他不能说话，

and I want to see what's stressing him out."

我想要知道什么东西会给他压力。”

And I said, "Sure, in fact, don't just take one, take two,"

我说到： “当然可以，实际上， 不要拿一个，拿两个吧”

because they broke easily back then.

因为它们那时很容易坏掉。

So he took them home, he put them on his little brother.

所以他把它们带回家， 并给他弟弟穿戴上了。

Now, I was back in MIT, looking at the data on my laptop,

然后，我回到了 MIT，看着我笔记本上的数据

and the first day, I thought, "Hmm, that's odd,

第一天，我想到： “额，这个有些古怪”

he put them on both wrists instead of waiting for one to break.

“他把传感器戴在了两个手腕上， 而不是等一个先坏掉。

OK, fine, don't follow my instructions."

不过没事，不用遵守我的指令。”

I'm glad he didn't.

我很欣慰他没有按照我的指令。

Second day -- chill. Looked like classroom activity.

第二天 -- 数据平稳低迷， 看起来像课堂活动。

(Laughter)

（大笑）

A few more days ahead.

又过去了几天。

The next day, one wrist signal was flat

一天，一个手腕的信号变平直了

and the other had the biggest peak I've ever seen,

而另外一个手腕的信号 出现了我从未遇到过的峰值。

and I thought, "What's going on?

我当时在想，“到底发生了什么？

We've stressed people out at MIT every way imaginable.

“我们在 MIT 给人们 各种可以想象到的压力。

I've never seen a peak this big."

我从来没见到过这么大的峰值。”

And it was only on one side.

而且它还只是一边的数据。

How can you be stressed on one side of your body and not the other?

你的身体怎样才能做到一边有压力 而不是另外一边呢？

So I thought one or both sensors must be broken.

所以，我以为其中一个或者 两个传感器一定是坏了。

Now, I'm an electroengineer by training,

我是一个受过培训的电子工程师，

so I started a whole bunch of stuff to try to debug this,

所以，我尝试了一堆方法来调试它，

and long story short, I could not reproduce this.

简短的说，我不能重现这个现象。

So I resorted to old-fashioned debugging.

所以，我又诉诸于老式的排障技术。

I called the student at home on vacation.

我给那个在家休假的学生挂了电话。

"Hi, how's your little brother? How's your Christmas?

“你好，你的弟弟情况如何？ 你的圣诞节怎么样？”

Hey, do you have any idea what happened to him?"

你弟弟发生了什么事吗？”

And I gave this particular date and time,

我告诉他了这个特殊的日期和时间

and the data.

还有当时的数据。

And he said, "I don't know, I'll check the diary."

他说到： “我不知道， 我会检查日记的”

Diary? An MIT student keeps a diary?

日记 ？ 一个 MIT 学生 会保持写日记？

So I waited and he came back.

所以，我等待着， 他一会儿就回话。

He had the exact date and time,

他告诉我精确的日期和时间，

and he says, "That was right before he had a grand mal seizure."

并且他说到： “那正是在他癫痫严重发作之前。”

Now, at the time, I didn't know anything about epilepsy,

在那时，我对癫痫一无所知，

and did a bunch of research,

然后我做了很多研究，

realized that another student's dad is chief of neurosurgery

而后我了解到另外一个学生的父亲是

at Children's Hospital Boston,

波士顿儿童医院的首席神经外科医生，

screwed up my courage and called Dr. Joe Madsen.

我鼓足勇气，打电话给乔·马森医生。

"Hi, Dr. Madsen, my name's Rosalind Picard.

“你好，乔·马森医生， 我叫罗莎琳德·皮卡尔。

Is it possible somebody could have

有没有可能某个人会产生

a huge sympathetic nervous system surge" --

巨大的交叉神经系统激增” --

that's what drives the skin conductance --

这就是驱动表层电导率激增的原因 --

"20 minutes before a seizure?"

“并且这会在癫痫发作前 20分钟出现 ？”

And he says, "Probably not."

他说： “可能不会。”

He says, "It's interesting.

他说： “这很有趣。

We've had people whose hair stands on end on one arm

我们有病人出现一只手臂的汗毛 竖起来的情况

20 minutes before a seizure."

在癫痫发作前20分钟。“

And I'm like, "On one arm?"

我说：“一只胳膊？”

I didn't want to tell him that, initially,

起初，我并不想告诉他那件事情，

because I thought this was too ridiculous.

因为我认为这太荒谬了。

He explained how this could happen in the brain,

他解释到这种情况如何在大脑中发生，

and he got interested. I showed him the data.

他也很感兴趣，我给他展示了数据。

We made a whole bunch more devices, got them safety certified.

我们制作了更多的设备， 并且通过安全认证。

90 families were being enrolled in a study,

90 个家庭参与到这个研究中，

all with children who were going to be monitored 24-7

这些家庭的孩子每时每刻都会 被监视测量，

with gold-standard EEG on their scalp

就通过装置于他们头皮上的 配备标准的脑电图分析器

for reading the brain activity,

来读取脑部活动，

video to watch the behavior,

还有视频来观察行为。

electrocardiogram -- ECG -- and now EDA, electrodermal activity,

心电监护仪还有皮肤电活动等

to see if there was something in this periphery

用来检查在这范围之内 是否存在某物

that we could easily pick up, related to a seizure.

使得我们能很快的发现， 并且它和癫痫有关。

We found, in 100 percent of the first batch of grand mal seizures,

我们发现，在第一批的癫痫大发作中， 100%的情况

this whopper of responses in the skin conductance.

都有这种皮肤电导率的巨大波动。

The blue in the middle, the boy's sleep,

中间的蓝色，男孩的睡眠，

is usually the biggest peak of the day.

通常是一天中的最大高峰值。

These three seizures you see here are popping out of the forest

这里看到的三个癫痫 就像从森林里冒出来

like redwood trees.

的红杉树。

Furthermore, when you couple the skin conductance at the top

另外，如果你把顶部的表皮电导率与

with the movement from the wrist

腕部的运动连接起来，

and you get lots of data and train machine learning and AI on it,

加上你得到的大量数据， 并且训练这方面的机器学习和AI，

you can build an automated AI that detects these patterns

你就能构建一个自动的人工智能 来更好地监测这些数据变动

much better than just a shake detector can do.

远比一个震动探测器能做的多。

So we realized that we needed to get this out,

我们意识到我们需要把它做出来，

and with the PhD work of Ming-Zher Poh

随着博明哲博士的工作

and later great improvements by Empatica,

以及后来Empatica公司的巨大改进后，

this has made progress and the seizure detection is much more accurate.

这个传感器取得了进步， 而且对癫痫的检测也变的更加精确。

But we also learned some other things about SUDEP during this.

另外在这个过程中， 我们又学到了更多关于SUDEP的知识。

One thing we learned is that SUDEP,

其中一个是，

while it's rare after a generalized tonic-clonic seizure,

SUDEP虽然很少会在 强直阵痉挛发生后

that's when it's most likely to happen -- after that type.

这是最有可能发生的时刻 -- 在那种类型之后。

And when it happens, it doesn't happen during the seizure,

SUDEP 并不在癫痫发作时发生，

and it doesn't usually happen immediately afterwards,

并且它通常不会马上发生，

but immediately afterwards,

但是，紧接着，

when the person just seems very still and quiet,

当病人看起来非常安稳时，

they may go into another phase, where the breathing stops,

他们可能会进入另外一个阶段， 他们会停止呼吸

and then after the breathing stops, later the heart stops.

在呼吸停止后，后来心脏停止。

So there's some time to get somebody there.

所以，存在一段时间 可以让某人赶过去。

We also learned that there is a region deep in the brain called the amygdala,

我们还了解到大脑深处 存在一个区域叫作杏仁核区

which we had been studying in our emotion research a lot.

我们在情感实验室 一直对它做了许多研究。

We have two amygdalas,

我们有两个杏仁核，

and if you stimulate the right one,

如果你刺激右边那一个

you get a big right skin conductance response.

你就会得到一个巨大的 右侧皮层传导率反应。

Now, you have to sign up right now for a craniotomy to get this done,

现在，你需要进行穿颅手术 来做到这一点

not exactly something we're going to volunteer to do,

这并不是什么我们会自愿去做的事情，

but it causes a big right skin conductance response.

但是它的确会引起一个巨大的 右侧表皮传导率反应。

Stimulate the left one, big left skin conductance response on the palm.

刺激左侧那个， 会引起手掌上左侧的皮层传导率反应。

And furthermore, when somebody stimulates your amygdala

更进一步说， 当某人刺激你的杏仁核时，

while you're sitting there and you might just be working,

你可能正坐着工作。

you don't show any signs of distress,

你并不会表现出来任何悲痛的信号

but you stop breathing,

但你会停止呼吸。

and you don't start again until somebody stimulates you.

并且在某人再次刺激你之前 你不会恢复呼吸。

"Hey, Roz, are you there?"

“你好，罗兹，你还好吧？”

And you open your mouth to talk.

这时你张开口开始说话。

As you take that breath to speak,

当你为了说话而吸了那口气之后，

you start breathing again.

你开始继续呼吸。

So we had started with work on stress,

所以，我们开始研究压力，

which had enabled us to build lots of sensors

这使得我们可以建造许多传感器，

that were gathering high quality enough data

它们能收集足够多的高质量数据，

that we could leave the lab and start to get this in the wild;

使得我们可以收集实验室之外的数据。

accidentally found a whopper of a response with the seizure,

我们无意间还发现了 癫痫引发的高峰反应

neurological activation that can cause a much bigger response

神经系统激活， 相对于传统的压力来说，

than traditional stressors;

能够引起更大的反应。

lots of partnership with hospitals and an epilepsy monitoring unit,

许多医院正与癫痫监控单元合作，

especially Children's Hospital Boston

特别是波士顿儿童医院

and the Brigham;

和布里格姆医院。

and machine learning and AI on top of this

现在还可利用机器学习和 人工智能的技术

to take and collect lots more data

用来获取和收集更多的数据

in service of trying to understand these events

以便尝试理解这些事件，

and if we could prevent SUDEP.

看看我们是否能够阻止 SUDEP。

This is now commercialized by Empatica,

Empatica 公司已将这些商业化

a start-up that I had the privilege to cofound,

这是一家初创企业， 我有幸是它的联合创办人，

and the team there has done an amazing job improving the technology

他们团队出色地改进了

to make a very beautiful sensor

制造更完美的传感器的技术，

that not only tells time and does steps and sleep and all that good stuff,

这个传感器不止告诉我们时间， 步数，睡眠这些好数据，

but this is running real-time AI and machine learning

它还实时运行着一个AI和 机器学习程序

to detect generalized tonic-clonic seizures

用来检测常规的强直阵痉挛型癫痫，

and send an alert for help

并会在我要发生癫痫并且失去意识时

if I were to have a seizure and lose consciousness.

发出警报。

This just got FDA-approved

这个已经被食药监局批准

as the first smartwatch to get approved in neurology.

作为第一个在神经系统方面 被批准的智能腕表。

(Applause)

（掌声）

Now, the next slide is what made my skin conductance go up.

现在，接下来的这张幻灯片 导致我的皮层传导率上升。

One morning, I'm checking my email

一天，我检查我的邮箱

and I see a story from a mom

我看到了一封来自一个妈妈的故事。

who said she was in the shower,

她说她正在淋浴，

and her phone was on the counter by the shower,

她的手机放在淋浴房的柜台上，

and it said her daughter might need her help.

手机告诉她女儿可能需要她的帮助。

So she interrupts her shower and goes running to her daughter's bedroom,

所以她中断洗澡奔跑进女儿的卧室

and she finds her daughter facedown in bed, blue and not breathing.

她发现女儿在被窝里脸朝下， 发紫并且没有呼吸。

She flips her over -- human stimulation --

她把她翻转过来 ——人类刺激

and her daughter takes a breath, and another breath,

然后她女儿吸了一口气， 接着另外一口气，

and her daughter turns pink and is fine.

最终她女儿脸色变粉并且好了。

I think I turned white reading this email.

我想我在读到这封邮件时脸色变白了。

My first response is, "Oh no, it's not perfect.

我的第一反应是 “噢，不，这还不完善。

The Bluetooth could break, the battery could die.

“蓝牙可能会坏掉，电池会耗尽。

All these things could go wrong. Don't rely on this."

这些都可能出错。你不能依赖这个。“

And she said, "It's OK. I know no technology is perfect.

然后她说 “没关系的， 我知道没有技术是完美的。

None of us can always be there all the time.

我们没有人可以一直在患者身边。

But this, this device plus AI

但是这个设备加上AI，

enabled me to get there in time to save my daughter's life."

让我能够及时挽救我女儿的性命。”

Now, I've been mentioning children,

现在，我一直在谈论小孩。

but SUDEP peaks, actually, among people in their 20s, 30s and 40s,

但是实际上，SUDEP 在20岁，30岁和 40岁人群中达到峰值。

and the next line I'm going to put up

还有下一个我将要展示的曲线

is probably going to make some people uncomfortable,

很可能会令一些人感到不舒适，

but it's less uncomfortable than we'll all be

但这其中若包括你们认识的人，

if this list is extended to somebody you know.

你们可能会更不适。

Could this happen to somebody you know?

这可能发生在你认识的某个人身上吗？

And the reason I bring up this uncomfortable question

我提出这个令人不适的问题的原因

is because one in 26 of you will have epilepsy at some point,

是因为你们当中每26个人 就有1个将会在某个点发生癫痫，

and from what I've been learning,

并且根据我所学到的，

people with epilepsy often don't tell their friends and their neighbors

患有癫痫的人通常不会 告诉他们的朋友和邻居

that they have it.

他们患有这种病的。

So if you're willing to let them use an AI or whatever

所以如果你愿意让他们 使用一个AI或者其它东西

to summon you in a moment of possible need,

在急需时刻召唤你的话，

if you would let them know that,

如果你让他们知道这些，

you could make a difference in their life.

你就可以改变他们的生活。

Why do all this hard work to build AIs?

为什么做这么多辛苦工作 来构建这些AI？

A couple of reasons here:

有几个原因：

one is Natasha, the girl who lived,

一个就是娜塔莎，那个活下来的女孩，

and her family wanted me to tell you her name.

她的家人想要我告诉你们她的名字。

Another is her family

另外一个是她的家庭

and the wonderful people out there

还有世界各地美好的人们，

who want to be there to support people who have conditions

他们想去帮助那些患有癫痫

that they've felt uncomfortable in the past mentioning to others.

却在过去不敢告诉他人的人们。

And the other reason is all of you,

另外一个原因就是你们所有人。

because we have the opportunity to shape the future of AI.

因为我们有机会去塑造 AI 的未来。

We can actually change it,

我们可以改变它

because we are the ones building it.

因为我们就是建设它的人。

So let's build AI

所以，让我们来建设 AI

that makes everybody's lives better.

来让每一个人的生活更好一些。

Thank you.

谢谢。

(Applause)

（掌声）