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

Filmed:
1,296,740 views

Parkinson's disease affects 6.3 million people worldwide, causing weakness and tremors, but there's no objective way to detect it early on. Yet. Applied mathematician and TED Fellow Max Little is testing a simple, cheap tool that in trials is able to detect Parkinson's with 99 percent accuracy -- in a 30-second phone call.
- 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 math,
0
667
2103
00:18
and this is a peculiar problem
1
2770
1524
00:20
for anyone who does applied math, is that
2
4294
2173
00:22
we are like management consultants.
3
6467
1933
00:24
No one knows what the hell we do.
4
8400
1946
00:26
So I am going to give you some -- attempt today
5
10346
2274
00:28
to try and explain to you what I do.
6
12620
2293
00:30
So, dancing is one of the most human of activities.
7
14913
3321
00:34
We delight at ballet virtuosos and tap dancers
8
18234
3682
00:37
you will see later on.
9
21916
1148
00:38
Now, ballet requires an extraordinary level of expertise
10
23064
2690
00:41
and a high level of skill,
11
25754
2914
00:44
and probably a level of initial suitability
12
28668
2531
00:47
that may well have a genetic component to it.
13
31199
1847
00:48
Now, sadly, neurological disorders such as Parkinson's disease
14
33046
3393
00:52
gradually destroy this extraordinary ability,
15
36439
2087
00:54
as it is doing to my friend Jan Stripling, who was
16
38526
2323
00:56
a virtuoso ballet dancer in his time.
17
40849
2967
00:59
So great progress and treatment has been made over the years.
18
43816
3054
01:02
However, there are 6.3 million people worldwide
19
46870
2944
01:05
who have the disease, and they have to live with
20
49814
3448
01:09
incurable weakness, tremor, rigidity
21
53262
2568
01:11
and the other symptoms that go along with the disease,
22
55830
1857
01:13
so what we need are objective tools
23
57687
2383
01:15
to detect the disease before it's too late.
24
60070
3057
01:19
We need to be able to measure progression objectively,
25
63127
2554
01:21
and ultimately, the only way we're going to know
26
65681
3173
01:24
when we actually have a cure is when we have
27
68854
2192
01:26
an objective measure that can answer that for sure.
28
71046
3398
01:30
But frustratingly, with Parkinson's disease
29
74444
2850
01:33
and other movement disorders, there are no biomarkers,
30
77294
2353
01:35
so there's no simple blood test that you can do,
31
79647
2229
01:37
and the best that we have is like
32
81876
1802
01:39
this 20-minute neurologist test.
33
83678
2241
01:41
You have to go to the clinic to do it. It's very, very costly,
34
85919
2458
01:44
and that means that, outside the clinical trials,
35
88377
2757
01:47
it's just never done. It's never done.
36
91134
2728
01:49
But what if patients could do this test at home?
37
93862
3077
01:52
Now, that would actually save on a difficult trip to the clinic,
38
96939
2098
01:54
and what if patients could do that test themselves, right?
39
99037
4254
01:59
No expensive staff time required.
40
103291
1920
02:01
Takes about $300, by the way,
41
105211
1418
02:02
in the neurologist's clinic to do it.
42
106629
1993
02:04
So what I want to propose to you as an unconventional way
43
108622
2681
02:07
in which we can try to achieve this,
44
111303
1514
02:08
because, you see, in one sense, at least,
45
112817
1808
02:10
we are all virtuosos like my friend Jan Stripling.
46
114625
3256
02:13
So here we have a video of the vibrating vocal folds.
47
117881
3755
02:17
Now, this is healthy and this is somebody making speech sounds,
48
121636
3229
02:20
and we can think of ourselves as vocal ballet dancers,
49
124865
3464
02:24
because we have to coordinate all of these vocal organs
50
128329
2214
02:26
when we make sounds, and we all actually
51
130543
2295
02:28
have the genes for it. FoxP2, for example.
52
132838
2296
02:31
And like ballet, it takes an extraordinary level of training.
53
135134
2713
02:33
I mean, just think how long it takes a child to learn to speak.
54
137847
2585
02:36
From the sound, we can actually track
55
140432
2382
02:38
the vocal fold position as it vibrates,
56
142814
2281
02:40
and just as the limbs are affected in Parkinson's,
57
145095
2543
02:43
so too are the vocal organs.
58
147638
2781
02:46
So on the bottom trace, you can see an example of
59
150419
1880
02:48
irregular vocal fold tremor.
60
152299
1698
02:49
We see all the same symptoms.
61
153997
1168
02:51
We see vocal tremor, weakness and rigidity.
62
155165
2930
02:53
The speech actually becomes quieter and more breathy
63
158095
2104
02:56
after a while, and that's one of the example symptoms of it.
64
160199
2233
02:58
So these vocal effects can actually be quite subtle,
65
162432
2847
03:01
in some cases, but with any digital microphone,
66
165279
3216
03:04
and using precision voice analysis software
67
168495
2545
03:06
in combination with the latest in machine learning,
68
171040
2409
03:09
which is very advanced by now,
69
173449
1578
03:10
we can now quantify exactly where somebody lies
70
175027
2886
03:13
on a continuum between health and disease
71
177913
2881
03:16
using voice signals alone.
72
180794
2596
03:19
So these voice-based tests, how do they stack up against
73
183390
2314
03:21
expert clinical tests? We'll, they're both non-invasive.
74
185704
2150
03:23
The neurologist's test is non-invasive. They both use existing infrastructure.
75
187854
3982
03:27
You don't have to design a whole new set of hospitals to do it.
76
191836
3004
03:30
And they're both accurate. Okay, but in addition,
77
194840
2302
03:33
voice-based tests are non-expert.
78
197142
3327
03:36
That means they can be self-administered.
79
200469
1992
03:38
They're high-speed, take about 30 seconds at most.
80
202461
2580
03:40
They're ultra-low cost, and we all know what happens.
81
205041
2294
03:43
When something becomes ultra-low cost,
82
207335
2440
03:45
it becomes massively scalable.
83
209775
2296
03:47
So here are some amazing goals that I think we can deal with now.
84
212071
3675
03:51
We can reduce logistical difficulties with patients.
85
215746
2426
03:54
No need to go to the clinic for a routine checkup.
86
218172
2312
03:56
We can do high-frequency monitoring to get objective data.
87
220484
2320
03:58
We can perform low-cost mass recruitment for clinical trials,
88
222804
4105
04:02
and we can make population-scale screening
89
226909
2115
04:04
feasible for the first time.
90
229024
1596
04:06
We have the opportunity to start to search
91
230620
2202
04:08
for the early biomarkers of the disease before it's too late.
92
232822
3541
04:12
So, taking the first steps towards this today,
93
236363
2758
04:15
we're launching the Parkinson's Voice Initiative.
94
239121
2126
04:17
With Aculab and PatientsLikeMe, we're aiming
95
241247
2232
04:19
to record a very large number of voices worldwide
96
243479
1928
04:21
to collect enough data to start to tackle these four goals.
97
245407
3140
04:24
We have local numbers accessible to three quarters
98
248547
1700
04:26
of a billion people on the planet.
99
250247
1610
04:27
Anyone healthy or with Parkinson's can call in, cheaply,
100
251857
3077
04:30
and leave recordings, a few cents each,
101
254934
2139
04:32
and I'm really happy to announce that we've already hit
102
257073
2190
04:35
six percent of our target just in eight hours.
103
259263
3543
04:38
Thank you. (Applause)
104
262806
3751
04:42
(Applause)
105
266557
6320
04:48
Tom Rielly: So Max, by taking all these samples of,
106
272877
3575
04:52
let's say, 10,000 people,
107
276452
2776
04:55
you'll be able to tell who's healthy and who's not?
108
279228
2854
04:57
What are you going to get out of those samples?
109
282082
1685
04:59
Max Little: Yeah. Yeah. So what will happen is that,
110
283767
1830
05:01
during the call you have to indicate whether or not
111
285597
1657
05:03
you have the disease or not, you see. TR: Right.
112
287254
1267
05:04
ML: You see, some people may not do it. They may not get through it.
113
288521
2507
05:06
But we'll get a very large sample of data that is collected
114
291028
2717
05:09
from all different circumstances, and it's getting it
115
293745
3408
05:13
in different circumstances that matter because then
116
297153
1905
05:14
we are looking at ironing out the confounding factors,
117
299058
3384
05:18
and looking for the actual markers of the disease.
118
302442
2161
05:20
TR: So you're 86 percent accurate right now?
119
304603
2497
05:23
ML: It's much better than that.
120
307100
1194
05:24
Actually, my student Thanasis, I have to plug him,
121
308294
1720
05:25
because he's done some fantastic work,
122
310014
1870
05:27
and now he has proved that it works over the mobile telephone network as well,
123
311884
3770
05:31
which enables this project, and we're getting 99 percent accuracy.
124
315654
3390
05:34
TR: Ninety-nine. Well, that's an improvement.
125
319044
1576
05:36
So what that means is that people will be able to —
126
320620
2201
05:38
ML: (Laughs)
127
322821
1852
05:40
TR: People will be able to call in from their mobile phones
128
324673
1906
05:42
and do this test, and people with Parkinson's could call in,
129
326579
3072
05:45
record their voice, and then their doctor can check up
130
329651
2870
05:48
on their progress, see where they're doing in this course of the disease.
131
332521
2681
05:51
ML: Absolutely.
132
335202
970
05:52
TR: Thanks so much. Max Little, everybody.
133
336172
1743
05:53
ML: Thanks, Tom. (Applause)
134
337915
5157
Translated by Joseph Geni
Reviewed by Morton Bast

▲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