ABOUT THE SPEAKER
Max Little - Applied mathematician
Max Little is a mathematician whose research includes a breakthrough technique to monitor – and potentially screen for – Parkinson's disease through simple voice recordings.

Why you should listen

Max Little is an applied mathematician whose goal is to "see connections between subjects, not boundaries … to see how things are related, not how they are different." He has a background in applied mathematics, statistics, signal processing and computational engineering, and his work has been applied across disciplines like biomedicine, extreme rainfall analysis and forecasting, biophysical signal processing, and hydrogeomorphology and open channel flow measurement. Little is best known for his work on the Parkinson's Voice Initiative, in which he and his team developed a cheap and simple tool that uses precise voice analysis software to detect Parkinson's with 99 percent accuracy. Little is a TEDGlobal 2012 Fellow and a Wellcome Trust-MIT Postdoctoral Research Fellow.

More profile about the speaker
Max Little | Speaker | TED.com
TEDGlobal 2012

Max Little: A test for Parkinson's with a phone call

Max Little:电话也能检测帕金森症

Filmed:
1,296,740 views

帕金森症影响了全球630万的人,使人们虚弱和僵化,但在早期却没有客观的方法对此进行监测。但是,应用数学家, TED伙伴 Max Little在测试一项简单廉价的方法,这在检测帕金森症的试验中,能达到99%的准确率——仅仅在一个30秒的来电就能搞定。
- Applied mathematician
Max Little is a mathematician whose research includes a breakthrough technique to monitor – and potentially screen for – Parkinson's disease through simple voice recordings. Full bio

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

00:16
So, well, I do applied应用的 math数学,
0
667
2103
话说,我是学应用数学的
00:18
and this is a peculiar奇特 problem问题
1
2770
1524
对于每一个学应用数学的人来说
00:20
for anyone任何人 who does applied应用的 math数学, is that
2
4294
2173
有一个特殊的难题,就是
00:22
we are like management管理 consultants顾问.
3
6467
1933
我们和管理顾问都很像
00:24
No one knows知道 what the hell地狱 we do.
4
8400
1946
没人知道我们是干嘛的
00:26
So I am going to give you some -- attempt尝试 today今天
5
10346
2274
所以我今天会给你们一些实例
00:28
to try and explain说明 to you what I do.
6
12620
2293
来解释我们是干什么的
00:30
So, dancing跳舞 is one of the most human人的 of activities活动.
7
14913
3321
嗯,跳舞是人类众多的活动之一
00:34
We delight at ballet芭蕾舞 virtuosos艺术大师 and tap龙头 dancers舞者
8
18234
3682
我们会因芭蕾舞大师和踢踏舞蹈者而感到欣悦
00:37
you will see later后来 on.
9
21916
1148
稍后你们就会知道了
00:38
Now, ballet芭蕾舞 requires要求 an extraordinary非凡 level水平 of expertise专门知识
10
23064
2690
如今芭蕾需要非凡的专业知识
00:41
and a high level水平 of skill技能,
11
25754
2914
和杰出的专业技巧
00:44
and probably大概 a level水平 of initial初始 suitability适应性
12
28668
2531
甚至一定的初级适应性
00:47
that may可能 well have a genetic遗传 component零件 to it.
13
31199
1847
这可能就是所说的天赋
00:48
Now, sadly可悲的是, neurological神经 disorders障碍 such这样 as Parkinson's帕金森氏 disease疾病
14
33046
3393
不过,可悲的是,神经系统疾病,例如帕金森症
00:52
gradually逐渐 destroy破坏 this extraordinary非凡 ability能力,
15
36439
2087
慢慢地在摧毁这一卓越的才能
00:54
as it is doing to my friend朋友 Jan一月 Stripling瘦长, who was
16
38526
2323
这正发生在我的朋友Jan Striling身上
00:56
a virtuoso演奏家 ballet芭蕾舞 dancer舞蹈家 in his time.
17
40849
2967
他曾是他那个时代的芭蕾舞大师
00:59
So great progress进展 and treatment治疗 has been made制作 over the years年份.
18
43816
3054
这些年他接受过很多治疗也有很大的进步
01:02
However然而, there are 6.3 million百万 people worldwide全世界
19
46870
2944
然而,世上有630万的人
01:05
who have the disease疾病, and they have to live生活 with
20
49814
3448
患有这种疾病,而且他们还得忍受
01:09
incurable不可救药 weakness弱点, tremor, rigidity刚性
21
53262
2568
无法治愈的虚弱,震颤,僵化
01:11
and the other symptoms症状 that go along沿 with the disease疾病,
22
55830
1857
等其他伴随而来的症状的折磨
01:13
so what we need are objective目的 tools工具
23
57687
2383
所以我们需要客观的工具
01:15
to detect检测 the disease疾病 before it's too late晚了.
24
60070
3057
在病重之前将之检测出来
01:19
We need to be able能够 to measure测量 progression级数 objectively客观地,
25
63127
2554
我们需要客观地测量进展
01:21
and ultimately最终, the only way we're going to know
26
65681
3173
最后,想要知道何时有治愈之方的唯一方法
01:24
when we actually其实 have a cure治愈 is when we have
27
68854
2192
就是当我们
01:26
an objective目的 measure测量 that can answer回答 that for sure.
28
71046
3398
拥有了能准确回答那些问题的客观测量法
01:30
But frustratingly令人沮丧, with Parkinson's帕金森氏 disease疾病
29
74444
2850
但令人泄气的是,帕金森症
01:33
and other movement运动 disorders障碍, there are no biomarkers生物标记物,
30
77294
2353
和其他运动障碍疾病,都没有生物标志物
01:35
so there's no simple简单 blood血液 test测试 that you can do,
31
79647
2229
所以你没法做简单的血液检查
01:37
and the best最好 that we have is like
32
81876
1802
我们能做到的最好的
01:39
this 20-minute-分钟 neurologist神经学家 test测试.
33
83678
2241
就是像这个20分钟的神经学测试
01:41
You have to go to the clinic诊所 to do it. It's very, very costly昂贵,
34
85919
2458
你必须得到诊所去,而且花费相当相当大
01:44
and that means手段 that, outside the clinical临床 trials试验,
35
88377
2757
这也意味着,在临床试验之外
01:47
it's just never doneDONE. It's never doneDONE.
36
91134
2728
从来没人做过这个测试,从来没有
01:49
But what if patients耐心 could do this test测试 at home?
37
93862
3077
但如果病人能在家完成这个测试呢?
01:52
Now, that would actually其实 save保存 on a difficult trip to the clinic诊所,
38
96939
2098
这也能为病人节省下那段去诊所的路程
01:54
and what if patients耐心 could do that test测试 themselves他们自己, right?
39
99037
4254
如果病人能独立完成这个测试呢?
01:59
No expensive昂贵 staff员工 time required需要.
40
103291
1920
就不需要花大量的钱去雇佣员工
02:01
Takes about $300, by the way,
41
105211
1418
顺便提下,如果去神经科诊所做的话
02:02
in the neurologist's神经学家的 clinic诊所 to do it.
42
106629
1993
这个费用要300美金
02:04
So what I want to propose提出 to you as an unconventional非传统的 way
43
108622
2681
所以我想向你们提出一个非常规的的方法
02:07
in which哪一个 we can try to achieve实现 this,
44
111303
1514
一个我们能尝试去达到目的的方法
02:08
because, you see, in one sense, at least最小,
45
112817
1808
因为你也知道,在某种意义上,至少
02:10
we are all virtuosos艺术大师 like my friend朋友 Jan一月 Stripling瘦长.
46
114625
3256
我们都是像我朋友Jan Stripling一样的艺术大师
02:13
So here we have a video视频 of the vibrating振动 vocal声音的 folds褶皱.
47
117881
3755
我这里有一个关于声襞颤动的视频
02:17
Now, this is healthy健康 and this is somebody making制造 speech言语 sounds声音,
48
121636
3229
这个是健康状态下的某人在发声
02:20
and we can think of ourselves我们自己 as vocal声音的 ballet芭蕾舞 dancers舞者,
49
124865
3464
我们把自己想当作是声音的芭蕾舞者
02:24
because we have to coordinate坐标 all of these vocal声音的 organs器官
50
128329
2214
因为当我们发声时
02:26
when we make sounds声音, and we all actually其实
51
130543
2295
我们得去协调这些发声器官,确实
02:28
have the genes基因 for it. FoxPFoxP2, for example.
52
132838
2296
这些是与生俱来的,例如Foxp2基因
02:31
And like ballet芭蕾舞, it takes an extraordinary非凡 level水平 of training训练.
53
135134
2713
而且和芭蕾一样,这也需要经过特殊地训练
02:33
I mean, just think how long it takes a child儿童 to learn学习 to speak说话.
54
137847
2585
可以想一想,小孩学说话得花多长时间
02:36
From the sound声音, we can actually其实 track跟踪
55
140432
2382
从声音里,我们可以追寻到
02:38
the vocal声音的 fold position位置 as it vibrates振动,
56
142814
2281
声襞颤动时的位置
02:40
and just as the limbs四肢 are affected受影响 in Parkinson's帕金森氏,
57
145095
2543
正如帕金森能影响我们四肢
02:43
so too are the vocal声音的 organs器官.
58
147638
2781
它也能影响我们发声器官
02:46
So on the bottom底部 trace跟踪, you can see an example of
59
150419
1880
你可以看到底部的颤动轨迹
02:48
irregular不规则 vocal声音的 fold tremor.
60
152299
1698
是一种不规则的声襞颤动
02:49
We see all the same相同 symptoms症状.
61
153997
1168
我们看到了一样的症状
02:51
We see vocal声音的 tremor, weakness弱点 and rigidity刚性.
62
155165
2930
声音颤动,虚弱,和僵化
02:53
The speech言语 actually其实 becomes quieter安静 and more breathy喘息
63
158095
2104
讲话会变得小声而且之后伴随更多气息音
02:56
after a while, and that's one of the example symptoms症状 of it.
64
160199
2233
这些都是症状的例子
02:58
So these vocal声音的 effects效果 can actually其实 be quite相当 subtle微妙,
65
162432
2847
事实上,这些声效大部分时候挺难以捉摸的
03:01
in some cases, but with any digital数字 microphone麦克风,
66
165279
3216
但只要有任何一款数码麦克风
03:04
and using运用 precision精确 voice语音 analysis分析 software软件
67
168495
2545
搭上精密的声音分析软件
03:06
in combination组合 with the latest最新 in machine learning学习,
68
171040
2409
再组合用上最新的目前已非常先进的
03:09
which哪一个 is very advanced高级 by now,
69
173449
1578
机器研究
03:10
we can now quantify量化 exactly究竟 where somebody lies
70
175027
2886
现在我们就能准确地定量
03:13
on a continuum连续 between之间 health健康 and disease疾病
71
177913
2881
他人处于健康和疾病这一直线上的位置
03:16
using运用 voice语音 signals信号 alone单独.
72
180794
2596
这仅仅只需要语音信号而已
03:19
So these voice-based基于语音的 tests测试, how do they stack up against反对
73
183390
2314
那么这些以语音为基础的测试,他们会怎样
03:21
expert专家 clinical临床 tests测试? We'll, they're both non-invasive无创.
74
185704
2150
与专业的临床测试展开较量呢?其实他们都是无创技术
03:23
The neurologist's神经学家的 test测试 is non-invasive无创. They both use existing现有 infrastructure基础设施.
75
187854
3982
神经学测试是无创性的,而且他们都使用现有的基础设备
03:27
You don't have to design设计 a whole整个 new set of hospitals医院 to do it.
76
191836
3004
你无需去设计一系列新的医院来完成测试
03:30
And they're both accurate准确. Okay, but in addition加成,
77
194840
2302
他们也都具有精确性。虽然这么说,但是
03:33
voice-based基于语音的 tests测试 are non-expert非专家.
78
197142
3327
基于语音的测试却不是专家级别的
03:36
That means手段 they can be self-administered自我管理.
79
200469
1992
这意味着患者能自己搞定
03:38
They're high-speed高速, take about 30 seconds at most.
80
202461
2580
而且速度也快,最多只要30秒
03:40
They're ultra-low超低 cost成本, and we all know what happens发生.
81
205041
2294
价格很低,我们也知道是怎么一回儿事
03:43
When something becomes ultra-low超低 cost成本,
82
207335
2440
一旦价格低廉
03:45
it becomes massively大规模 scalable可扩展性.
83
209775
2296
它也就能被大规模地扩展
03:47
So here are some amazing惊人 goals目标 that I think we can deal合同 with now.
84
212071
3675
我这儿有些现在能拿出来说的伟大的目标
03:51
We can reduce减少 logistical后勤 difficulties困难 with patients耐心.
85
215746
2426
我们能减少和病人有关的后勤问题
03:54
No need to go to the clinic诊所 for a routine常规 checkup检查.
86
218172
2312
没必要去诊所做例行的检查
03:56
We can do high-frequency高频 monitoring监控 to get objective目的 data数据.
87
220484
2320
我们能进行高频率的监测以获取客观数据
03:58
We can perform演出 low-cost低成本 mass recruitment招聘 for clinical临床 trials试验,
88
222804
4105
我们能为临床试验组织大量征召且耗费很低
04:02
and we can make population-scale人口规模 screening筛查
89
226909
2115
我们也能使人数的规模的展示
04:04
feasible可行 for the first time.
90
229024
1596
在第一时间变得可行
04:06
We have the opportunity机会 to start开始 to search搜索
91
230620
2202
我们也有机会开始寻找
04:08
for the early biomarkers生物标记物 of the disease疾病 before it's too late晚了.
92
232822
3541
这种疾病在病入膏肓前的早期生物标志物
04:12
So, taking服用 the first steps脚步 towards this today今天,
93
236363
2758
所以,从今天开始朝这个方向迈出第一步
04:15
we're launching发射 the Parkinson's帕金森氏 Voice语音 Initiative倡议.
94
239121
2126
我们开始了Parkinson's Voice Initiative(帕金森主创声音测试)这个项目
04:17
With AculabAculab公司 and PatientsLikeMePatientsLikeMe, we're aiming瞄准
95
241247
2232
和Aculab还有PatientsLikeMe一起合作,我们希望
04:19
to record记录 a very large number of voices声音 worldwide全世界
96
243479
1928
记录下大量世界各地的声音
04:21
to collect搜集 enough足够 data数据 to start开始 to tackle滑车 these four goals目标.
97
245407
3140
收集足够的数据来达到这四个目标
04:24
We have local本地 numbers数字 accessible无障碍 to three quarters住处
98
248547
1700
我们的市话号码能接通地球上
04:26
of a billion十亿 people on the planet行星.
99
250247
1610
7亿5000万人的来电
04:27
Anyone任何人 healthy健康 or with Parkinson's帕金森氏 can call in, cheaply廉价地,
100
251857
3077
任何健康或者患有帕金森疾病的人都能打进来,很便宜
04:30
and leave离开 recordings录音, a few少数 cents each,
101
254934
2139
留下录音,每次只要几分钱
04:32
and I'm really happy快乐 to announce宣布 that we've我们已经 already已经 hit击中
102
257073
2190
我很高兴宣布我们仅在8小时内
04:35
six percent百分 of our target目标 just in eight hours小时.
103
259263
3543
就达到了我们目标的6%
04:38
Thank you. (Applause掌声)
104
262806
3751
谢谢(掌声)
04:42
(Applause掌声)
105
266557
6320
(掌声)
04:48
Tom汤姆 Rielly里利: So Max马克斯, by taking服用 all these samples样本 of,
106
272877
3575
Tom Rielly: Max,靠着这些采样
04:52
let's say, 10,000 people,
107
276452
2776
不妨说,在一万人中
04:55
you'll你会 be able能够 to tell who's谁是 healthy健康 and who's谁是 not?
108
279228
2854
你能够辨别出谁是健康的,谁是患病的?
04:57
What are you going to get out of those samples样本?
109
282082
1685
你又想从这些采样中得出什么呢?
04:59
Max马克斯 Little: Yeah. Yeah. So what will happen发生 is that,
110
283767
1830
Max Little:对,是的。所以是这样的
05:01
during the call you have to indicate表明 whether是否 or not
111
285597
1657
在来电期间,你得告诉我们
05:03
you have the disease疾病 or not, you see. TRTR: Right.
112
287254
1267
你是否患有这种疾病。TR:嗯
05:04
MLML: You see, some people may可能 not do it. They may可能 not get through通过 it.
113
288521
2507
ML:要知道,有些人可能不会打给我们,或者没接通
05:06
But we'll get a very large sample样品 of data数据 that is collected
114
291028
2717
但我们还是会得到大量
05:09
from all different不同 circumstances情况, and it's getting得到 it
115
293745
3408
来自不同环境下的采样数据
05:13
in different不同 circumstances情况 that matter because then
116
297153
1905
而且重点是不同的环境,因为这样
05:14
we are looking at ironing熨烫 out the confounding混杂 factors因素,
117
299058
3384
我们才能消除掉那些混淆的因素
05:18
and looking for the actual实际 markers标记 of the disease疾病.
118
302442
2161
寻求这种疾病真正的标记
05:20
TRTR: So you're 86 percent百分 accurate准确 right now?
119
304603
2497
TR:所以你现在拥有86%的准确率了?
05:23
MLML: It's much better than that.
120
307100
1194
ML:已经远远不止了
05:24
Actually其实, my student学生 ThanasisThanasis, I have to plug插头 him,
121
308294
1720
事实上,我的学生Thanasis,我得提一下他
05:25
because he's doneDONE some fantastic奇妙 work,
122
310014
1870
因为他做出一些出色的成果
05:27
and now he has proved证实 that it works作品 over the mobile移动 telephone电话 network网络 as well,
123
311884
3770
现在他已经证明了在移动电话网上也能进行这个项目
05:31
which哪一个 enables使 this project项目, and we're getting得到 99 percent百分 accuracy准确性.
124
315654
3390
所以现在我们接近99%的准确率了
05:34
TRTR: Ninety-nine九十九. Well, that's an improvement起色.
125
319044
1576
TR:九十九,好吧,这还挺了不起的
05:36
So what that means手段 is that people will be able能够 to —
126
320620
2201
所以这也意味着人们能——
05:38
MLML: (Laughs)
127
322821
1852
ML:(笑)
05:40
TRTR: People will be able能够 to call in from their mobile移动 phones手机
128
324673
1906
TR:人们能用手机打进你们的电话
05:42
and do this test测试, and people with Parkinson's帕金森氏 could call in,
129
326579
3072
然后做这项测试,帕金森病人也能打通电话
05:45
record记录 their voice语音, and then their doctor医生 can check up
130
329651
2870
然后记录下声音,之后他们的医生就能检查他们的进展
05:48
on their progress进展, see where they're doing in this course课程 of the disease疾病.
131
332521
2681
确定他们处于这种疾病的哪个阶段
05:51
MLML: Absolutely绝对.
132
335202
970
ML:就是这样
05:52
TRTR: Thanks谢谢 so much. Max马克斯 Little, everybody每个人.
133
336172
1743
TR:非常感谢,Max Little和所有人
05:53
MLML: Thanks谢谢, Tom汤姆. (Applause掌声)
134
337915
5157
ML:谢谢你Tom(掌声)
Translated by Lee Li
Reviewed by Yuguo Zhang

▲Back to top

ABOUT THE SPEAKER
Max Little - Applied mathematician
Max Little is a mathematician whose research includes a breakthrough technique to monitor – and potentially screen for – Parkinson's disease through simple voice recordings.

Why you should listen

Max Little is an applied mathematician whose goal is to "see connections between subjects, not boundaries … to see how things are related, not how they are different." He has a background in applied mathematics, statistics, signal processing and computational engineering, and his work has been applied across disciplines like biomedicine, extreme rainfall analysis and forecasting, biophysical signal processing, and hydrogeomorphology and open channel flow measurement. Little is best known for his work on the Parkinson's Voice Initiative, in which he and his team developed a cheap and simple tool that uses precise voice analysis software to detect Parkinson's with 99 percent accuracy. Little is a TEDGlobal 2012 Fellow and a Wellcome Trust-MIT Postdoctoral Research Fellow.

More profile about the speaker
Max Little | Speaker | TED.com