An AI smartwatch that detects seizures | Rosalind Picard

譯者: Lilian Chiu 審譯者: Helen Chang

This is Henry,

這位是亨利，

a cute boy,

一位可愛的男孩，

and when Henry was three,

亨利三歲時，

his mom found him having some febrile seizures.

他的媽媽發現他有點熱痙攣。

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

熱痙攣就是在痙攣的 同時還加上發燒，

and the doctor said,

而醫生說：

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

「不用太擔心。通常孩子 長大之後就沒有了。」

When he was four, he had a convulsive seizure,

他四歲時出現了痙攣性癲癇，

the kind that you lose consciousness and shake --

會失去意識且發抖的那一種——

a generalized tonic-clonic seizure --

全身強直陣攣型發作——

and while the diagnosis of epilepsy was in the mail,

當癲癇診斷書還在郵寄的途中，

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

一天早上亨利的媽媽去叫他起床，

and as she went in his room,

她進入他的房間，

she found his cold, lifeless body.

她發現他的身體冰冷， 沒有生命跡象。

Henry died of SUDEP,

亨利死於 SUDEP，

sudden unexpected death in epilepsy.

即癲癇猝死症。

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

我很好奇，在座有多少人 聽過癲癇猝死症？

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

各位是教育程度很高的觀眾， 但我只看到幾個人舉手。

SUDEP is when an otherwise healthy person with epilepsy

癲癇猝死症指的是 健康的癲癇患者

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

在死亡解剖後無法發現任何死因。

There is a SUDEP every seven to nine minutes.

每七到九分鐘就有一個人 發生癲癇猝死症。

That's on average two per TED Talk.

平均每場 TED 演說就有兩個人。

Now, a normal brain has electrical activity.

正常人的大腦都會有電訊活動。

You can see some of the electrical waves

各位可以在這張大腦的圖片上

coming out of this picture of a brain here.

看到一些電波。

And these should look like typical electrical activity

這些是腦電圖從表面讀到的

that an EEG could read on the surface.

典型腦電活動。

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

當你發作時，就會有 不尋常的腦電活動，

and it can be focal.

這可能就是病灶所在。

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

它可能只發生在大腦的一塊小區域。

When that happens, you might have a strange sensation.

發生時，你可能會有 很奇怪的感覺。

Several could be happening here in the audience right now,

在座當中可能就有幾個人 現在就有這個狀況，

and the person next to you might not even know.

而你旁邊的人可能完全不會知道。

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

然而，如果你腦中的發作狀況 是從小小的火苗散播到

like a forest fire over the brain,

像是森林大火，

then it generalizes,

它就會變成全身性，

and that generalized seizure takes your consciousness away

而全身性的發作會奪去你的意識，

and causes you to convulse.

造成你抽搐。

There are more SUDEPs in the United States every year

在美國，每年癲癇猝死的人數

than sudden infant death syndrome.

高於嬰兒猝死症。

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

在座有多少人聽過嬰兒猝死症？

Right? Pretty much every hand goes up.

是吧？幾乎人人都舉手了。

So what's going on here?

這是怎麼回事？

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

為什麼這種疾病這麼普遍， 然而大家卻沒聽說過它呢？

And what can you do to prevent it?

還有，要如何預防它呢？

Well, there are two things, scientifically shown,

嗯，科學指出，有兩個方法

that prevent or reduce the risk of SUDEP.

可以預防或減低 癲癇猝死症的風險。

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

第一：「遵循醫生指示，

take your medications."

按時吃藥。」

Two-thirds of people who have epilepsy

三分之二的癲癇患者

get it under control with their medications.

都能用藥物來控制。

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

能減低癲癇猝死症風險的 第二個方法，就是陪伴。

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

也就是當你發作的時候， 要有人在你身邊。

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

雖然大部分人沒聽過癲癇猝死症，

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

但它其實是所有神經疾病當中，

of all neurological disorders.

最有可能致死的第二名。

The vertical axis is the number of deaths

縱軸是死亡人數，

times the remaining life span,

乘上剩下的生命期長，

so higher is much worse impact.

所以越高表示越糟糕的影響。

SUDEP, however, unlike these others,

然而，癲癇猝死症不像其他疾病，

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

這裡的人都能夠 做點什麼來壓制住它。

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

為什麼人工智慧研究者 羅莎琳皮卡德

要在這裡跟大家談癲癇猝死症？

I'm not a neurologist.

我不是神經病學家。

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

我在麻省理工媒體實驗室 研究情緒測量時，

trying to make our machines more intelligent about our emotions,

我試圖要讓我們的機器 有更高的智能來判讀我們的情緒，

we started doing a lot of work measuring stress.

我們開始努力去測量壓力。

We built lots of sensors

我們建造了許多感測器，

that measured it in lots of different ways.

用許多不同的方式來測量壓力。

But one of them in particular

但，其中有一種特別方式，

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

出自用電訊號來測量掌心流汗的

with an electrical signal.

古老研究。

This is a signal of skin conductance

這是皮膚電導的訊號，

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

已知當你緊張時，這種訊號會升高，

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

但，結果發現，還有許多其他 有趣的條件也會讓它上升。

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

但，手上連結電線來測量 實在不太方便。

所以，我們在麻省理工 媒體實驗室裡

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

發明了許多其他方法來做這件事。

And with these wearables,

有了這些可穿戴的配備，

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

我們開始能取得一週七天、 一天二十四小時的

高品質臨床資料， 這是前所未有的。

Here's a picture of what that looked like

這張圖上的是

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

第一次從一位麻省理工學院 學生的手腕收集到

一週七天、一天二十四小時的 皮膚電導資料。

Let's zoom in a little bit here.

咱們把它放大一點。

What you see is 24 hours from left to right,

從左到右是二十四小時，

and here is two days of data.

圖上的是兩天份的資料。

And first, what surprised us

首先，讓我們驚訝的是，

was sleep was the biggest peak of the day.

一天的高峰發生在睡眠時間。

Now, that sounds broken, right?

聽起來是故障了，對吧？

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

你沉睡時是很冷靜的， 所以這是怎麼回事？

Well, it turns out that our physiology during sleep

結果發現，在睡眠期間的生理機能

is very different than our physiology during wake,

和清醒期間非常不同，

and while there's still a bit of a mystery

雖然對於為什麼一天之中的高峰

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

通常出現在睡眠期間， 這仍然有點難以理解，

we now believe they're related to memory consolidation

但我們現在相信，它們和記憶穩固

and memory formation during sleep.

及睡眠期間的記憶形成有關。

We also saw things that were exactly what we expected.

也有一些發現， 完全在我們的預期之中。

When an MIT student is working hard in the lab

當麻省理工學院學生 在實驗室中努力或在做功課時，

or on homeworks,

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

不僅有情緒壓力，還有認知負荷，

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

結果發現，認知負荷、 認知努力、心理參與、

excitement about learning something --

對學習某件事物感到興奮——

those things also make the signal go up.

上述這些也都會造成訊號上升。

Unfortunately, to the embarrassment of we MIT professors,

不幸的是，讓我們麻省理工學院 教授們很丟臉的是，

(Laughter)

（笑聲）

the low point every day is classroom activity.

一天的低點出現在教室活動期間。

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

各位在這裡看到的， 只是一個人的資料，

but this, unfortunately, is true in general.

但，不幸的是，一般來說 大家都有這個現象。

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

這條防汗帶裡面 有自製的皮膚電導感測器，

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

有一天，我們的一位 大學生來敲我的門，

right at the end of the December semester,

那時是十二月學期末，

and he said, "Professor Picard,

他說：「皮卡德教授，

can I please borrow one of your wristband sensors?

我能不能跟你借一個腕帶感測器？

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

我弟弟有自閉症，他無法說話，

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

我想了解是什麼讓他很焦慮。」

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

我說：「當然， 事實上，拿兩條去吧。」

because they broke easily back then.

因為在那時， 這種感測器很容易壞。

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

他把感測器帶回家， 讓他弟弟戴上。

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

而我在麻省理工學院裡， 用我的筆記型電腦看資料，

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

第一天，我心想：「嗯，真怪，

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

他把兩個感測器戴在兩隻手腕上， 而不是等一個壞了。

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

好，不照我的指示就算了。」

I'm glad he didn't.

我很高興他沒照我的指示。

Second day -- chill. Looked like classroom activity.

第二天——冷淡。 看起來就像教室活動期間。

(Laughter)

（笑聲）

A few more days ahead.

還要觀察幾天。

The next day, one wrist signal was flat

隔天，有一隻手腕的訊號很平坦，

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

另一隻則有我見過最高的高峰，

and I thought, "What's going on?

我心想：「這是怎麼回事？

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

我們在麻省理工學院裡用過 各種想得出來的方式製造焦慮。

I've never seen a peak this big."

我從沒見過這麼高的高峰。」

And it was only on one side.

且只發生在一隻手。

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

怎麼可能身體的一邊 很焦慮，另一邊卻不會？

So I thought one or both sensors must be broken.

所以我認為，至少 其中一個感測器壞了。

Now, I'm an electroengineer by training,

我是受過訓練的電機工程師，

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

所以我開始動手去偵錯，

and long story short, I could not reproduce this.

長話短說，我無法複製這個狀況。

So I resorted to old-fashioned debugging.

所以，我回去用老式的偵錯方法。

I called the student at home on vacation.

這位學生放假在家， 我打電話給他：

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

「嗨，你弟弟如何？ 聖誕節過得如何？

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

嘿，你知道他發生了什麼事嗎？」

And I gave this particular date and time,

我提出了這個特定的日期、 時間，以及資料。

and the data.

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

他說：「我不知道， 我來查一下日記。」

Diary? An MIT student keeps a diary?

日記？麻省理工學院的 學生會寫日記？

So I waited and he came back.

我等待著，他回來了。

He had the exact date and time,

他有精確的日期和時間，

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

他說：「那剛好就是他 癲癇大發作之前。」

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

當時，我對癲癇一無所知，

and did a bunch of research,

於是我做了一堆研究，

realized that another student's dad is chief of neurosurgery

我發現有一個學生的爸爸是

波士頓兒童醫院的神經外科主任，

at Children's Hospital Boston,

screwed up my courage and called Dr. Joe Madsen.

我鼓起勇氣，打電話給 喬馬德森醫生。

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

「嗨，馬德森醫生， 我是羅莎琳皮卡德。

Is it possible somebody could have

有沒有可能，有人會發生

a huge sympathetic nervous system surge" --

交感神經系統突然高漲」——

that's what drives the skin conductance --

皮膚電導就是這麼來的——

"20 minutes before a seizure?"

「然後二十分鐘後就發作？」

And he says, "Probably not."

他說：「應該不會。」

He says, "It's interesting.

他說：「這很有趣。

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

我們見過病人一隻手臂上的毛聳立，

20 minutes before a seizure."

然後二十分鐘後就發作。」

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

我說：「一隻手臂？」 一開始，我還不想告訴他這件事，

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

because I thought this was too ridiculous.

因為我覺得太荒謬了。

He explained how this could happen in the brain,

他解釋了大腦中 可能發生了什麼事，

and he got interested. I showed him the data.

他很感興趣，我把資料給他看。

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

我們又做了一大堆裝置， 且為它們取得安全認證。

90 families were being enrolled in a study,

一項研究招募了九十個家庭參加，

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

他們的孩子會受到一週七天， 一天二十四小時的監控，

with gold-standard EEG on their scalp

頭皮會接上高品質的腦波儀， 來讀取大腦活動數據，

for reading the brain activity,

video to watch the behavior,

有影片可觀察行為，

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

還有心電圖—— ECG —— 現在是看 EDA，膚電活動，

to see if there was something in this periphery

用來尋找周圍是否有什麼我們可以

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

很容易取得和發作有關的訊號。

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

我們發現，

第一批的癲癇大發作

this whopper of responses in the skin conductance.

全部都伴隨這種超大的膚電反應。

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

中間的藍色區段是男孩的 睡眠期間，通常都是一天的高峰。

is usually the biggest peak of the day.

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

現在看到的這三次發作特別突出，

like redwood trees.

就像森林中的巨杉。

Furthermore, when you couple the skin conductance at the top

此外，如果把上面的膚電資料

with the movement from the wrist

搭配來自手腕的活動，

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

就會得到很多資料，可以用來 訓練機器學習和人工智慧，

you can build an automated AI that detects these patterns

可以建造自動化人工智慧， 來偵測這些模式，

much better than just a shake detector can do.

比晃動偵測器能做的要多太多。

So we realized that we needed to get this out,

我們知道我們得要 把這些成果傳出去，

and with the PhD work of Ming-Zher Poh

加上傅明哲的博士論文研究

and later great improvements by Empatica,

和 Empatica 公司後續做的大改良，

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

讓我們有了進展， 發作偵測的精確度大大提升。

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

過程中，我們也更進一步 了解了癲癇猝死症。

One thing we learned is that SUDEP,

關於癲癇猝死症， 我們學到的一件事就是，

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

雖然很少會在全身 強直陣攣型發作之後，

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

那是它最可能發生的時段—— 在那種發作之後。

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

當它發生時，並不會 在發作的過程中發生，

and it doesn't usually happen immediately afterwards,

通常不會緊接著馬上發生，

but immediately afterwards,

但，緊接著，

when the person just seems very still and quiet,

當病人還看似非常安穩的時候，

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

可能就會進入另一個階段， 在這個階段中，呼吸停止，

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

在呼吸停止後，心臟接著停止。

So there's some time to get somebody there.

所以還要花些時間才會到那一步。

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

我們也發現，在大腦深處 有一個區域叫做杏仁核，

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

情緒研究常常會去探究這個區域。

We have two amygdalas,

我們有兩個杏仁核， 如果去刺激右邊的那個，

and if you stimulate the right one,

you get a big right skin conductance response.

產生的右側膚電反應就會很大。

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

要願意接受顱骨切開術 才能做到這件事。

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

我們不會想要自願去做，

but it causes a big right skin conductance response.

但它會造成很大的右側膚電反應。

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

刺激左邊的，手掌會有 很大的左側膚電反應。

And furthermore, when somebody stimulates your amygdala

此外，當有人刺激你的杏仁核，

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

當時你坐在那裡，可能正在工作，

you don't show any signs of distress,

你不會顯露出任何苦難的徵兆，

but you stop breathing,

但你會停止呼吸，

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

而且除非有人去刺激你， 你不會再次呼吸。

"Hey, Roz, are you there?"

「嘿，羅莎，你在嗎？」

And you open your mouth to talk.

你打開你的嘴巴說話。

As you take that breath to speak,

當你吸入空氣以便說話，

you start breathing again.

你又開始呼吸了。

So we had started with work on stress,

所以，我們開始研究壓力，

which had enabled us to build lots of sensors

這讓我們能夠打造許多感測器，

that were gathering high quality enough data

用來收集品質夠高的資料，

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

能夠收集實驗室以外的資料；

accidentally found a whopper of a response with the seizure,

無意間發現發作帶來的高峰反應，

neurological activation that can cause a much bigger response

這種神經系統活化，造成的反應 遠大於傳統的壓力因子；

than traditional stressors;

lots of partnership with hospitals and an epilepsy monitoring unit,

和許多醫院及一個 癲癇監控單位合作，

especially Children's Hospital Boston

特別是波士頓兒童醫院， 及布萊根婦女醫院；

and the Brigham;

and machine learning and AI on top of this

此外還有機器學習和人工智慧，

to take and collect lots more data

來取得、收集更多資料，

in service of trying to understand these events

用來試圖了解這些事件，

and if we could prevent SUDEP.

希望能夠預防癲癇猝死症。

This is now commercialized by Empatica,

現在，Empatica 已經將它商業化，

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

它是間新創公司， 我很榮幸是共同創辦人，

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

那裡的團隊把這項技術 做了很驚人的改良，

to make a very beautiful sensor

做出一個很棒的感測器，

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

它不僅能夠報時， 也提供睡眠相關各種資訊，

but this is running real-time AI and machine learning

而且還有即時的 人工智慧和機器學習，

to detect generalized tonic-clonic seizures

可以偵測全身強直陣攣型發作，

and send an alert for help

並在我即將發作且失去意識時

if I were to have a seizure and lose consciousness.

發出求助警告。

This just got FDA-approved

它剛得到食品及藥物 管理局的核准，

as the first smartwatch to get approved in neurology.

是神經學領域中第一支 被核准的智慧手錶。

(Applause)

（掌聲）

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

下一張投影片呈現的是 我的皮膚電導上升的原因。

One morning, I'm checking my email

有天早上，我在讀電子郵件，

and I see a story from a mom

看到了一位母親的故事，

who said she was in the shower,

她說她當時正在淋浴，

and her phone was on the counter by the shower,

她的手機在淋浴間的檯子上，

and it said her daughter might need her help.

手機說她的女兒 可能需要她的協助。

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

所以她淋浴到一半 就衝到她女兒的臥房，

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

她發現她的女兒面朝下 趴在床上，發青，沒有呼吸。

She flips her over -- human stimulation --

她把她翻過來——人類刺激——

and her daughter takes a breath, and another breath,

她的女兒吸了一口氣， 接著再一口，

and her daughter turns pink and is fine.

她的女兒漸漸恢復血色，沒事了。

I think I turned white reading this email.

我想我在讀這封信時 我應該是臉色發白的。

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

我的第一個反應是： 「喔，不，它不夠完美。

The Bluetooth could break, the battery could die.

藍芽可能會斷掉，電池可能會用完。

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

這些都可能會出錯。 不要仰賴這個裝置。

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

她說：「沒關係，我知道 沒有任何技術是完美的。

None of us can always be there all the time.

沒有人能隨時都在身邊。

But this, this device plus AI

但，這個裝置加上人工智慧，

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

讓我能及時趕到， 救了我女兒一命。」

Now, I've been mentioning children,

我剛剛一直在談孩子，

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

但其實，癲癇猝死症最常發生在 二十、三十、四十多歲的人身上，

and the next line I'm going to put up

我接下來要說的，

is probably going to make some people uncomfortable,

可能會讓一些人不舒服，

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

但，如果這份名單 包括到你認識的人，

if this list is extended to somebody you know.

就會更不舒服。

Could this happen to somebody you know?

這有沒有可能發生在 你認識的人身上？

And the reason I bring up this uncomfortable question

我之所以提出這個不舒服的問題，

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

是因為二十六個人當中就有 一個人，會在某個時點罹患癲癇，

and from what I've been learning,

據我所知，

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

罹患癲癇的人通常 不會告訴朋友和鄰居。

that they have it.

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

所以，如果你願意讓他們 使用人工智慧之類的東西

to summon you in a moment of possible need,

在他們可能需要協助時呼叫你，

if you would let them know that,

如果你能讓他們知道這點，

you could make a difference in their life.

你就能讓他們的人生有所不同。

Why do all this hard work to build AIs?

為什麼要花這麼多功夫 來打造人工智慧？

A couple of reasons here:

這裡有幾個理由：

one is Natasha, the girl who lived,

第一個理由是娜塔莎， 活下來的那位女孩，

and her family wanted me to tell you her name.

她的家人要我把 她的名字告訴大家。

Another is her family

另一個理由是她的家人

and the wonderful people out there

以及其他美好的人，

who want to be there to support people who have conditions

他們想要支援那些 過去不敢告訴他人

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

自己患有某些疾病的人。

And the other reason is all of you,

還有一個理由是你們所有人，

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

因為我們有機會形塑 人工智慧的未來。

We can actually change it,

我們能夠改變它，

because we are the ones building it.

因為我們就是建造它的人。

So let's build AI

所以，咱們來打造出

that makes everybody's lives better.

能改善大家生活的人工智慧。

Thank you.

謝謝。

(Applause)

（掌聲）