How AI is making it easier to diagnose disease | Pratik Shah

87,253 views ใƒป 2018-08-21

TED


ืื ื ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ืœืžื˜ื” ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ.

ืชืจื’ื•ื: Talia Breuer ืขืจื™ื›ื”: Ido Dekkers
00:13
Computer algorithms today are performing incredible tasks
0
13280
3856
ื›ื™ื•ื, ืืœื’ื•ืจื™ืชืžื™ื ืžืžื•ื—ืฉื‘ื™ื ืžื‘ืฆืขื™ื ืคืขื•ืœื•ืช ืžื“ื”ื™ืžื•ืช
00:17
with high accuracies, at a massive scale, using human-like intelligence.
1
17160
4736
ื‘ืจืžืช ื“ื™ื•ืง ื’ื‘ื•ื”ื”, ื‘ื›ืžื•ื™ื•ืช ื ืจื—ื‘ื•ืช ื‘ืขื–ืจืช ื—ื™ืงื•ื™ ืฉืœ ื—ื›ืžืช ื‘ื ื™ ืื“ื.
00:21
And this intelligence of computers is often referred to as AI
2
21920
3936
ื•ื”ืดื—ื›ืžื”ืด ื”ื–ื• ืฉืœ ื”ืžื—ืฉื‘ื™ื ืžืชื•ืืจืช ืคืขืžื™ื ืจื‘ื•ืช ื›ื‘ืดืž
00:25
or artificial intelligence.
3
25880
1856
ืื• ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช.
00:27
AI is poised to make an incredible impact on our lives in the future.
4
27760
4200
ื‘ืดืž ืžื™ื•ืขื“ืช ืœื™ื™ืฆืจ ื”ืฉืคืขื” ืขืฆื•ืžื” ืขืœ ื”ื—ื™ื™ื ืฉืœื ื• ื‘ืขืชื™ื“.
00:32
Today, however, we still face massive challenges
5
32880
3936
ืœืžืจื•ืช ื–ืืช, ื”ื™ื•ื, ืื ื• ืขื“ื™ื™ืŸ ืžืชืžื•ื“ื“ื™ื ืขื ืืชื’ืจื™ื ื›ื‘ื“ื™ื
00:36
in detecting and diagnosing several life-threatening illnesses,
6
36840
3496
ื‘ืื™ืชื•ืจ ื•ืื‘ื—ื•ืŸ ืฉืœ ืžืกืคืจ ืžื—ืœื•ืช ืžืกื›ื ื•ืช ื—ื™ื™ื,
00:40
such as infectious diseases and cancer.
7
40360
2360
ื›ืžื• ืžื—ืœื•ืช ืžื“ื‘ืงื•ืช ื•ืกืจื˜ืŸ.
00:44
Thousands of patients every year
8
44000
2296
ืืœืคื™ ืžื˜ื•ืคืœื™ื ืžื“ื™ ืฉื ื”
00:46
lose their lives due to liver and oral cancer.
9
46320
2800
ืžืื‘ื“ื™ื ืืช ื—ื™ื™ื”ื ืขืงื‘ ืกืจื˜ืŸ ื”ื›ื‘ื“ ื•ื”ืคื”.
00:49
Our best way to help these patients
10
49880
2696
ื”ื“ืจืš ื”ื˜ื•ื‘ื” ื‘ื™ื•ืชืจ ืœืขื–ื•ืจ ืœืื•ืชื ื”ืžื˜ื•ืคืœื™ื
00:52
is to perform early detection and diagnoses of these diseases.
11
52600
4320
ื”ื™ื ืื™ืชื•ืจ ื•ืื‘ื—ื•ืŸ ืžื•ืงื“ื ืฉืœ ื”ืžื—ืœื•ืช ื”ืืœื•.
00:57
So how do we detect these diseases today, and can artificial intelligence help?
12
57880
4160
ืื– ื›ื™ืฆื“ ืื ื• ืžืืชืจื™ื ืืช ื”ืžื—ืœื•ืช ื”ืืœื• ื”ื™ื•ื, ื•ืื™ืš ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื™ื›ื•ืœื” ืœืขื–ื•ืจ?
01:03
In patients who, unfortunately, are suspected of these diseases,
13
63920
3656
ืœืžื˜ื•ืคืœื™ื ืืฉืจ, ืœืฆืขืจื ื• ื”ืจื‘, ืื ื• ื—ื•ืฉื“ื™ื ืฉื—ื•ืœื™ื ื‘ืื—ืช ืžืŸ ื”ืžื—ืœื•ืช,
01:07
an expert physician first orders
14
67600
2656
ืจื•ืคื ืžื•ืžื—ื” ื™ื–ืžื™ืŸ ื“ื‘ืจ ืจืืฉื•ืŸ
01:10
very expensive medical imaging technologies
15
70280
2616
ื”ื“ืžื™ื•ืช ืจืคื•ืื™ื•ืช ื‘ื˜ื›ื ื•ืœื•ื’ื™ื•ืช ื™ืงืจื•ืช ืžืื•ื“
01:12
such as fluorescent imaging, CTs, MRIs, to be performed.
16
72920
4096
ื›ื’ื•ืŸ ืจื ื˜ื’ืŸ ืคืœื•ืื•ืจื•ืกื ื˜ื™, ืกื™-ื˜ื™, ืืž-ืืจ-ืื™ื™.
01:17
Once those images are collected,
17
77040
2296
ื‘ืจื’ืข ืฉื”ื”ื“ืžื™ื•ืช ื ืืกืคื•ืช,
01:19
another expert physician then diagnoses those images and talks to the patient.
18
79360
4520
ืจื•ืคื ืžื•ืžื—ื” ื ื•ืกืฃ ืžืื‘ื—ืŸ ืืช ื”ื”ื“ืžื™ื•ืช ื•ืžืฉื•ื—ื— ืขื ื”ืžื˜ื•ืคืœ.
01:24
As you can see, this is a very resource-intensive process,
19
84520
3456
ื›ืคื™ ืฉืืชื ืจื•ืื™ื, ืžื“ื•ื‘ืจ ื‘ืชื”ืœื™ืš ื”ื“ื•ืจืฉ ืžืฉืื‘ื™ื ืจื‘ื™ื,
01:28
requiring both expert physicians, expensive medical imaging technologies,
20
88000
4416
ื‘ื™ื ื™ื”ื ืฉื ื™ ืจื•ืคืื™ื ืžื•ืžื—ื™ื, ื˜ื›ื ื•ืœื•ื’ื™ื•ืช ื”ื“ืžื™ื” ืจืคื•ืื™ืช ื™ืงืจื•ืช,
01:32
and is not considered practical for the developing world.
21
92440
3096
ื•ืชื”ืœื™ืš ื–ื” ืœื ื ื—ืฉื‘ ื™ืขื™ืœ ื‘ื™ื—ืก ืœืขื•ืœื ื”ืžืชืคืชื—.
01:35
And in fact, in many industrialized nations, as well.
22
95560
3360
ืœืžืขืฉื”, ืืคื™ืœื• ื‘ื™ื—ืก ืœื”ืจื‘ื” ืžื“ื™ื ื•ืช ืžืคื•ืชื—ื•ืช.
01:39
So, can we solve this problem using artificial intelligence?
23
99760
2880
ืื–, ื”ืื ื ื™ืชืŸ ืœืคืชื•ืจ ืืช ื”ื‘ืขื™ื” ื‘ืขื–ืจืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช?
01:43
Today, if I were to use traditional artificial intelligence architectures
24
103840
4056
ื›ื™ื•ื, ืื ื”ื™ื™ืชื™ ืžืฉืชืžืฉ ื‘ืฉื™ื˜ื•ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ืžืกื•ืจืชื™ื•ืช
01:47
to solve this problem,
25
107920
1216
ืœืคืชืจื•ืŸ ื”ื‘ืขื™ื”,
01:49
I would require 10,000 --
26
109160
1456
ื ื“ืจืฉื™ื ืœื™ 10,000--
01:50
I repeat, on an order of 10,000 of these very expensive medical images
27
110640
4016
ืื ื™ ื—ื•ื–ืจ, ืกื“ืจ ื’ื•ื“ืœ ืฉืœ 10,000 ื”ื“ืžื™ื•ืช ืจืคื•ืื™ื•ืช ื™ืงืจื•ืช
01:54
first to be generated.
28
114680
1376
ืœืฉืœื‘ ื”ืจืืฉื•ื ื™.
01:56
After that, I would then go to an expert physician,
29
116080
2896
ืœืื—ืจ ืžื›ืŸ, ืืฆื˜ืจืš ืœื”ืคื’ืฉ ืขื ืจื•ืคื ืžื•ืžื—ื”,
01:59
who would then analyze those images for me.
30
119000
2496
ืืฉืจ ื™ืฆื˜ืจืš ืœื ืชื— ืืช ื”ื”ื“ืžื™ื•ืช ื”ืืœื• ืขื‘ื•ืจื™.
02:01
And using those two pieces of information,
31
121520
2096
ื•ื‘ืขื–ืจืช ืฉืชื™ ืคื™ืกื•ืช ื”ืžื™ื“ืข ื”ืืœื•,
02:03
I can train a standard deep neural network or a deep learning network
32
123640
3656
ืื ื™ ื™ื›ื•ืœ ืœืืžืŸ ืจืฉืช ื ื•ื™ืจื•ื ื™ื ืกื˜ื ื“ืจื˜ื™ืช ืื• ืจืฉืช ืœืžื™ื“ื”
02:07
to provide patient's diagnosis.
33
127320
2136
ืขืœ ืœืžื ืช ืœืกืคืง ืื‘ื—ื ื” ืœืžื˜ื•ืคืœ.
02:09
Similar to the first approach,
34
129480
1736
ื‘ื“ื•ืžื” ืœื’ื™ืฉื” ื”ืจืืฉื•ื ื”,
02:11
traditional artificial intelligence approaches
35
131240
2143
ื’ื™ืฉื•ืช ืžืกื•ืจืชื™ื•ืช ืœื‘ื™ื ื” ืžืœืื›ื•ืชื™ืชึฟ
02:13
suffer from the same problem.
36
133407
1449
ืกื•ื‘ืœื•ืช ืžืื•ืชื” ื”ื‘ืขื™ื”.
02:14
Large amounts of data, expert physicians and expert medical imaging technologies.
37
134880
4560
ื›ืžื•ื™ื•ืช ื’ื“ื•ืœื•ืช ืฉืœ ืžื™ื“ืข, ืจื•ืคืื™ื ืžื•ืžื—ื™ื ื•ื˜ื›ื ื•ืœื•ื’ื™ื•ืช ื”ื“ืžื™ื” ืจืคื•ืื™ืช ืžืชื•ื—ื›ืžื•ืช.
02:20
So, can we invent more scalable, effective
38
140320
4296
ืžื›ืืŸ, ื”ืื ื ื™ืชืŸ ืœื”ืžืฆื™ื ืคืชืจื•ืŸ ืžื“ืจื’ื™, ื™ืขื™ืœ
02:24
and more valuable artificial intelligence architectures
39
144640
3296
ื•ืœื™ืฆื•ืจ ืžืขืจื›ื•ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื‘ืขืœื•ืช ืขืจืš ื’ื“ื•ืœ ื™ื•ืชืจ
02:27
to solve these very important problems facing us today?
40
147960
3056
ืœืคืชืจื•ืŸ ื‘ืขื™ื•ืช ื—ืฉื•ื‘ื•ืช ืžืกื•ื’ ื–ื” ืขืžืŸ ืื ื• ืžืชืžื•ื“ื“ื™ื ื›ื™ื•ื?
02:31
And this is exactly what my group at MIT Media Lab does.
41
151040
3296
ื•ื–ื” ื‘ื“ื™ื•ืง ืžื” ืฉื”ืงื‘ื•ืฆื” ืฉืœื™ ืขื•ืฉื”.
02:34
We have invented a variety of unorthodox AI architectures
42
154360
3856
ื”ืžืฆืื ื• ืžื’ื•ื•ืŸ ืžืขืจื›ื•ืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ื—ื“ืฉื ื™ื•ืช
02:38
to solve some of the most important challenges facing us today
43
158240
3176
ืขืœ ืžื ืช ืœืคืชื•ืจ ื›ืžื” ืžื”ืืชื’ืจื™ื ื”ื—ืฉื•ื‘ื™ื ื‘ื™ื•ืชืจ ืขืžื ืื ื• ืžืชืžื•ื“ื“ื™ื ื›ื™ื•ื
02:41
in medical imaging and clinical trials.
44
161440
2200
ื‘ื”ื“ืžื™ื” ืจืคื•ืื™ืช ื•ื ื™ืกื•ื™ื™ื ืงืœื™ื ื™ื™ื.
02:44
In the example I shared with you today, we had two goals.
45
164480
3056
ื‘ืขื–ืจืช ื”ื“ื•ื’ืžื” ืฉืฉื™ืชืคืชื™ ืื™ืชื›ื ื”ื™ื•ื ื”ืฆื‘ื ื• ืฉืชื™ ืžื˜ืจื•ืช.
02:47
Our first goal was to reduce the number of images
46
167560
2976
ื”ืžื˜ืจื” ื”ืจืืฉื•ื ื” ื”ื™ื ืœืฆืžืฆื ืืช ืžืกืคืจ ื”ื”ื“ืžื™ื•ืช
02:50
required to train artificial intelligence algorithms.
47
170560
3256
ื”ื ื“ืจืฉื•ืช ืขืœ ืžื ืช ืœืืžืŸ ืืœื’ื•ืจื™ืชื ืฉืœ ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช.
02:53
Our second goal -- we were more ambitious,
48
173840
2096
ื”ืžื˜ืจื” ื”ืฉื ื™ื” -- ื”ื™ื™ืชื” ืœื ื• ื™ื•ืชืจ ืชืขื•ื–ื”,
02:55
we wanted to reduce the use of expensive medical imaging technologies
49
175960
3736
ืจืฆื™ื ื• ืœืฆืžืฆื ืืช ื”ืฉื™ืžื•ืฉ ื‘ื˜ื›ื ื•ืœื•ื’ื™ื•ืช ื”ื“ืžื™ื” ืจืคื•ืื™ืช ื™ืงืจื•ืช
02:59
to screen patients.
50
179720
1216
ื‘ืฉื‘ื™ืœ ืœืกื ืŸ ืžื˜ื•ืคืœื™ื.
03:00
So how did we do it?
51
180960
1200
ื›ื™ืฆื“ ืขืฉื™ื ื• ื–ืืช?
03:02
For our first goal,
52
182920
1216
ืขื‘ื•ืจ ื”ืžื˜ืจื” ื”ืจืืฉื•ื ื”,
03:04
instead of starting with tens and thousands
53
184160
2056
ื‘ืžืงื•ื ืœื”ืชื—ื™ืœ ื‘ืขืฉืจื•ืช ืืœืคื™ื
03:06
of these very expensive medical images, like traditional AI,
54
186240
3016
ืžืŸ ื”ื”ื“ืžื™ื•ืช ื”ื™ืงืจื•ืช ื”ืœืœื•, ื›ืžื• ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช ืžืกื•ืจืชื™ืช,
03:09
we started with a single medical image.
55
189280
2056
ื”ืชื—ืœื ื• ืžื”ื“ืžื™ื” ื‘ื•ื“ื“ืช.
03:11
From this image, my team and I figured out a very clever way
56
191360
3776
ื‘ืขื–ืจืช ืื•ืชื” ื”ื”ื“ืžื™ื”, ื”ืฆื•ื•ืช ื•ืื ื™ ืžืฆืื ื• ืฉื™ื˜ื” ื—ื›ืžื”
03:15
to extract billions of information packets.
57
195160
2736
ืœืืกื•ืฃ ื‘ื™ืœื™ื•ื ื™ื ืฉืœ ืžื ื•ืช ืžื™ื“ืข.
03:17
These information packets included colors, pixels, geometry
58
197920
3696
ืžื ื•ืช ื”ืžื™ื“ืข ื”ืืœื• ื›ืœืœื• ืฆื‘ืขื™ื, ืคื™ืงืกืœื™ื, ื’ืื•ืžื˜ืจื™ื”
03:21
and rendering of the disease on the medical image.
59
201640
2536
ื•ื”ืชืจื’ื•ื ืฉืœ ื”ืžื—ืœื” ืขืœ ื”ื”ื“ืžื™ื” ื”ืจืคื•ืื™ืช.
03:24
In a sense, we converted one image into billions of training data points,
60
204200
4336
ื‘ืฆื•ืจื” ืžืกื•ื™ืžืช, ื”ืžืจื ื• ื”ื“ืžื™ื” ืื—ืช ืœื‘ื™ืœื™ื•ื ื™ื ืฉืœ ื ืชื•ื ื™ื ืœืœืžื™ื“ื”,
03:28
massively reducing the amount of data needed for training.
61
208560
3536
ืชื•ืš ืฆืžืฆื•ื ืื“ื™ืจ ืฉืœ ื›ืžื•ืช ื”ืžื™ื“ืข ื”ื ื“ืจืฉ ืœืœืžื™ื“ื”.
03:32
For our second goal,
62
212120
1216
ืขื‘ื•ืจ ื”ืžื˜ืจื” ื”ืฉื ื™ื™ื”,
03:33
to reduce the use of expensive medical imaging technologies to screen patients,
63
213360
3856
ืฆืžืฆื•ื ืฉืœ ื”ืฉื™ืžื•ืฉ ื‘ื˜ื›ื ื•ืœื•ื’ื™ื•ืช ื”ื“ืžื™ื” ืจืคื•ืื™ืช ืœืกื™ื ื•ืŸ ืžื˜ื•ืคืœื™ื,
03:37
we started with a standard, white light photograph,
64
217240
2856
ื”ืชื—ืœื ื• ื‘ืฆื™ืœื•ื ืชืžื•ื ื” ืกื˜ื ื“ืจื˜ื™ืช,
03:40
acquired either from a DSLR camera or a mobile phone, for the patient.
65
220120
4336
ื”ื ืœืงื—ื” ืžืžืฆืœืžืช DSLR ืื• ื˜ืœืคื•ืŸ ืกืœื•ืœืจื™, ืขื‘ื•ืจ ื”ืžื˜ื•ืคืœ.
03:44
Then remember those billions of information packets?
66
224480
2456
ื•ืื–, ื–ื•ื›ืจื™ื ืืช ื‘ื™ืœื™ื•ื ื™ ืžื ื•ืช ื”ืžื™ื“ืข?
03:46
We overlaid those from the medical image onto this image,
67
226960
3536
ื”ื—ืœื ื• ืื•ืชืŸ ืžื”ื”ื“ืžื™ื” ื”ืจืคื•ืื™ืช ืขืœ ื’ื‘ื™ ื”ืชืžื•ื ื” ืฉืฆื™ืœืžื ื•,
03:50
creating something that we call a composite image.
68
230520
2520
ื•ื™ืฆืจื ื• ื“ื‘ืจ ืฉืงืจืื ื• ืœื• ืชืžื•ื ื” ืžืจื•ื›ื‘ืช.
03:53
Much to our surprise, we only required 50 --
69
233480
3296
ืœืžืจื‘ื” ื”ื”ืคืชืขื”, ื ื“ืจืฉื• ืœื ื• ืจืง 50--
03:56
I repeat, only 50 --
70
236800
1336
ืื ื™ ื—ื•ื–ืจ, ืจืง 50--
03:58
of these composite images to train our algorithms to high efficiencies.
71
238160
3840
ืžื”ืชืžื•ื ื•ืช ื”ืžืจื•ื›ื‘ื•ืช ื”ืœืœื• ืขืœ ืžื ืช ืœืืžืŸ ืืช ื”ืืœื’ื•ืจื™ืชืžื™ื ืฉืœื ื• ื‘ืจืžืช ื“ื™ื•ืง ื’ื‘ื•ื”ื”.
04:02
To summarize our approach,
72
242680
1336
ืœืกื™ื›ื•ื ื”ื’ื™ืฉื” ืฉืœื ื•,
04:04
instead of using 10,000 very expensive medical images,
73
244040
3176
ื‘ืžืงื•ื ืœื”ืฉืชืžืฉ ื‘10,000 ื”ื“ืžื™ื•ืช ืจืคื•ืื™ื•ืช ื™ืงืจื•ืช,
04:07
we can now train the AI algorithms in an unorthodox way,
74
247240
3016
ืื ื• ืžืกื•ื’ืœื™ื ื›ืขืช ืœืืžืŸ ืืœื’ื•ืจื™ืชืžื™ื ืฉืœ ื‘ืดืž ื‘ืฆื•ืจื” ืœื ืžืกื•ืจืชื™ืช,
04:10
using only 50 of these high-resolution, but standard photographs,
75
250280
4256
ืขืœ ื™ื“ื™ ืฉื™ืžื•ืฉ ื‘ 50 ืชืžื•ื ื•ืช ืกื˜ื ื“ืจื˜ื™ื•ืช ื‘ืื™ื›ื•ืช ื’ื‘ื•ื”ื”
04:14
acquired from DSLR cameras and mobile phones,
76
254560
2496
ืฉื ืœืงื—ื• ื‘ืขื–ืจืช ืžืฆืœืžื•ืช DSLR ื•ืžื›ืฉื™ืจื™ื ืกืœื•ืœืจื™ื™ื,
04:17
and provide diagnosis.
77
257080
1536
ื•ืœืกืคืง ืื‘ื—ื ื”.
04:18
More importantly,
78
258640
1216
ื—ืฉื•ื‘ ืžื›ืš,
04:19
our algorithms can accept, in the future and even right now,
79
259880
3136
ื”ืืœื’ื•ืจื™ืชืžื™ื ืฉืœื ื• ืžืกื•ื’ืœื™ื ืœืงืœื•ื˜, ื‘ืขืชื™ื“ ื•ืืคื™ืœื• ืขื›ืฉื™ื•,
04:23
some very simple, white light photographs from the patient,
80
263040
2816
ื’ื ืชืžื•ื ื•ืช ืคืฉื•ื˜ื•ืช ืฉืœ ื”ืžื˜ื•ืคืœ,
04:25
instead of expensive medical imaging technologies.
81
265880
2440
ื‘ืžืงื•ื ื”ื“ืžื™ื•ืช ืจืคื•ืื™ื•ืช ื‘ื˜ื›ื ื•ืœื•ื’ื™ื•ืช ื™ืงืจื•ืช.
04:29
I believe that we are poised to enter an era
82
269120
3096
ืื ื™ ืžืืžื™ืŸ ื›ื™ ืื ื• ืขื•ืžื“ื™ื ืœื”ื›ื ืก ืœืขื™ื“ืŸ
04:32
where artificial intelligence
83
272240
1936
ื‘ื• ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช
04:34
is going to make an incredible impact on our future.
84
274200
2536
ื”ื•ืœื›ืช ืœื™ื™ืฆืจ ื”ืฉืคืขื” ืขืฆื•ืžื” ืขืœ ื”ืขืชื™ื“ ืฉืœื ื•.
04:36
And I think that thinking about traditional AI,
85
276760
2456
ื•ืื ื™ ื—ื•ืฉื‘ ืฉืชื•ืš ื—ืฉื™ื‘ื” ืขืœ ื‘ืดืž ืžืกื•ืจืชื™ืช,
04:39
which is data-rich but application-poor,
86
279240
2776
ืืฉืจ ืขืฉื™ืจื” ื‘ื ืชื•ื ื™ื ืืš ื—ืœืฉื” ื‘ื™ื™ืฉื•ื,
04:42
we should also continue thinking
87
282040
1536
ืขืœื™ื ื• ืœื”ืžืฉื™ืš ืœื—ืฉื•ื‘
04:43
about unorthodox artificial intelligence architectures,
88
283600
3016
ืขืœ ืฉื™ื˜ื•ืช ื—ื“ืฉื ื™ื•ืช ืœื™ืฆื™ืจืช ื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช
04:46
which can accept small amounts of data
89
286640
1936
ืืฉืจ ื™ื›ื•ืœื•ืช ืœืขื‘ื•ื“ ืขื ื›ืžื•ืช ืงื˜ื ื” ืฉืœ ืžื™ื“ืข
04:48
and solve some of the most important problems facing us today,
90
288600
2936
ื•ืœืคืชื•ืจ ื›ืžื” ืžืŸ ื”ื‘ืขื™ื•ืช ื”ื—ืฉื•ื‘ื•ืช ื‘ื™ื•ืชืจ ืžื•ืœืŸ ืื ื• ืขื•ืžื“ื™ื ื”ื™ื•ื,
04:51
especially in health care.
91
291560
1256
ื‘ื™ื™ื—ื•ื“ ื‘ืจืคื•ืื”.
04:52
Thank you very much.
92
292840
1216
ืชื•ื“ื” ืจื‘ื” ืœื›ื.
04:54
(Applause)
93
294080
3840
(ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
ืขืœ ืืชืจ ื–ื”

ืืชืจ ื–ื” ื™ืฆื™ื’ ื‘ืคื ื™ื›ื ืกืจื˜ื•ื ื™ YouTube ื”ืžื•ืขื™ืœื™ื ืœืœื™ืžื•ื“ ืื ื’ืœื™ืช. ืชื•ื›ืœื• ืœืจืื•ืช ืฉื™ืขื•ืจื™ ืื ื’ืœื™ืช ื”ืžื•ืขื‘ืจื™ื ืขืœ ื™ื“ื™ ืžื•ืจื™ื ืžื”ืฉื•ืจื” ื”ืจืืฉื•ื ื” ืžืจื—ื‘ื™ ื”ืขื•ืœื. ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ื”ืžื•ืฆื’ื•ืช ื‘ื›ืœ ื“ืฃ ื•ื™ื“ืื• ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ ืžืฉื. ื”ื›ืชื•ื‘ื™ื•ืช ื’ื•ืœืœื•ืช ื‘ืกื ื›ืจื•ืŸ ืขื ื”ืคืขืœืช ื”ื•ื•ื™ื“ืื•. ืื ื™ืฉ ืœืš ื”ืขืจื•ืช ืื• ื‘ืงืฉื•ืช, ืื ื ืฆื•ืจ ืื™ืชื ื• ืงืฉืจ ื‘ืืžืฆืขื•ืช ื˜ื•ืคืก ื™ืฆื™ืจืช ืงืฉืจ ื–ื”.

https://forms.gle/WvT1wiN1qDtmnspy7