ABOUT THE SPEAKER
Pratik Shah - Medical technologist
Dr. Pratik Shah creates novel intersections between engineering, medical imaging, machine learning and medicine.

Why you should listen

Dr. Shah's research program at the MIT Media Lab develops scalable and low-cost diagnostics and therapeutics. His ongoing research areas at MIT include: 1) artificial intelligence and machine learning methods for detection of cancer biomarkers using standard photographs vs. expensive medical images; 2) unorthodox artificial intelligence and machine learning algorithms to design optimal and faster clinical trials and to reduce adverse effects on patients; and 3) low-cost and open source imaging devices, paper diagnostics, algorithms and mobile phones to improve public health and generate real-world data.

Clinical studies with Pratik's medical technologies have revealed "missing sick" patients, who otherwise remain undiagnosed in conventional healthcare settings. Dr. Shah's graduate and postdoctoral research contributed to the discovery of a vaccine component to prevent pneumococcal (Streptococcus pneumoniae) diseases; the identification of new pathways, technologies and metabolites as antimicrobials to target gastrointestinal infections; and a nonprofit to deploy a low-cost water quality test for the developing world.

Past recognition for Dr. Shah includes the American Society for Microbiology's Raymond W. Sarber national award, the Harvard Medical School and Massachusetts General Hospitals ECOR Fund for Medical Discovery postdoctoral fellowship, the AAAS-Lemelson Invention Ambassador Award and a TED Fellowship. Pratik has been an invited discussion leader at Gordon Research Seminars; a speaker at Cold Spring Harbor Laboratories, Gordon Research Conferences and IEEE bioengineering conferences; and a peer reviewer for leading scientific publications and funding agencies. Pratik has a BS, MS, and a PhD in Microbiology and completed fellowship training at The Broad Institute of MIT and Harvard, Massachusetts General Hospital and Harvard Medical School.

More profile about the speaker
Pratik Shah | Speaker | TED.com
TEDGlobal 2017

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

Pratik Shah: Kako umjetna inteligencija dijagnosticira oboljenja

Filmed:
1,571,835 views

Današnji algoritmi UI zahtijevaju hiljade skupih medicinskih snimaka kako bi otkrili oboljenje pacijenta. Šta ako bi mogli drastično smanjiti količinu podataka potrebnih za treniranje UI, čineći dijagnoze jeftinije i efikasnije? TED Fellow Pratik Shah radi na pametnom sistemu koji čini upravo to. Koristeći neuobičajen pristupi UI, Shah je razvio tehnologiju koja uz pomoć samo 50 slika razvija funkcionalan algoritam - i koji, također, može koristiti slike sa mobilnih uređaja kako bi postavili dijagnozu. Naučite kako ovaj novi način analiziranja medicinskih informacija može doprinijeti ranijem otkrivanju smrtonosnih oboljenja i primijeniti dijagnosticiranje pomoću UI u zdravstvu širom svijeta.
- Medical technologist
Dr. Pratik Shah creates novel intersections between engineering, medical imaging, machine learning and medicine. Full bio

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

00:13
ComputerRačunalo algorithmsalgoritmi todaydanas
are performingizvođenje incrediblenevjerovatno taskszadataka
0
1280
3856
Kompjuterski algoritmi čine
nevjerojatne stvari danas
00:17
with highvisoka accuraciespreciznosti, at a massivemasivni scaleskala,
usingkoristeći human-likeljudski intelligenceinteligencija.
1
5160
4736
sa visokim nivoom preciznosti,
koristeći inteligenciju sličnu ljudskoj.
00:21
And this intelligenceinteligencija of computersračunari
is oftenčesto referredupućeni to as AIAl
2
9920
3936
Ova kompjuterska inteligencija
se često naziva UI
00:25
or artificialveštački intelligenceinteligencija.
3
13880
1856
ili umjetna inteligencija.
00:27
AIAl is poisedstaložen to make an incrediblenevjerovatno impactuticaj
on our livesživi in the futurebudućnost.
4
15760
4200
UI je predodređena da ostavi veliki
utjecaj na naše živote.
00:32
TodayDanas, howeverkako god,
we still facelice massivemasivni challengesizazovi
5
20880
3936
Ipak, danas se susrećemo
sa velikim izazovima
00:36
in detectingotkrivanje and diagnosingdijagnozaciju
severalnekoliko life-threateningživotne prijetnje illnessesbolesti,
6
24840
3496
u otkrivanju i dijagnosticiranju
mnogih opasnih oboljenja,
00:40
suchtakve as infectiouszarazne diseasesbolesti and cancerrak.
7
28360
2360
kao što su zarazne bolesti i rak.
00:44
ThousandsTisuće of patientspacijenti everysvaki yeargodina
8
32000
2296
Hiljade pacijenata svake godine
00:46
losegube theirnjihova livesživi
duedue to liverjetra and oraloralni cancerrak.
9
34320
2800
izgube život zbog
raka jetre i usne šupljine.
00:49
Our bestnajbolje way to help these patientspacijenti
10
37880
2696
Najbolji način na koji možemo
pomoći ovim pacijentima
00:52
is to performizvoditi earlyrano detectionotkrivanje
and diagnosesdijagnoze of these diseasesbolesti.
11
40600
4320
je da ranije otkrijemo
i dijagnosticiramo ova oboljenja.
00:57
So how do we detectdetektovati these diseasesbolesti todaydanas,
and can artificialveštački intelligenceinteligencija help?
12
45880
4160
Pa kako možemo otkriti ta oboljenja danas,
i može li umjetna inteligencija pomoći?
01:03
In patientspacijenti who, unfortunatelynažalost,
are suspectedsumnjivo of these diseasesbolesti,
13
51920
3656
Nažalost, za pacijente koji sumnjaju
da boluju od ovih oboljenja
01:07
an expertstručnjak physicianliječnik first ordersnaredbe
14
55600
2656
stručnjaci prvo preporučuju
01:10
very expensiveskupo
medicalmedicinski imagingObrada slike technologiestehnologije
15
58280
2616
vrlo skupe tehnologije
zdravstvenog snimanja
01:12
suchtakve as fluorescentfluorescentne imagingObrada slike,
CTsAta, MRIsMRI, to be performedizvršeno.
16
60920
4096
kao flourescentna snimanja,
rendgene i magnetne rezonance.
01:17
OnceJednom those imagesslike are collectedprikupljeni,
17
65040
2296
Kada se ta snimanja obave,
01:19
anotherdrugi expertstručnjak physicianliječnik then diagnosesdijagnoze
those imagesslike and talksrazgovore to the patientpacijent.
18
67360
4520
drugi stručnjak dijagnosticira
te snimke i razgovara sa pacijentom.
01:24
As you can see, this is
a very resource-intensiveresursno-intenzivna processproces,
19
72520
3456
Kao što vidite, ovo je jako skup
i iscrpljujući proces,
01:28
requiringkoje zahtijevaju bothoboje expertstručnjak physicianslekari,
expensiveskupo medicalmedicinski imagingObrada slike technologiestehnologije,
20
76000
4416
koji zahtijeva i medicinske stručnjake
i skupe tehnologije snimanja,
01:32
and is not consideredrazmotrena practicalpraktično
for the developingrazvoj worldsvet.
21
80440
3096
te nije uopće praktičan
za zemlje u razvoju.
01:35
And in factčinjenica, in manymnogi
industrializedindustrijski nationsnacije, as well.
22
83560
3360
Odnosno, nije ni za
razvijenije zemlje, također.
01:39
So, can we solveriješiti this problemproblem
usingkoristeći artificialveštački intelligenceinteligencija?
23
87760
2880
Možemo li riješiti ovaj problem
umjetnom inteligencijom?
01:43
TodayDanas, if I were to use traditionaltradicionalno
artificialveštački intelligenceinteligencija architecturesarhitekte
24
91840
4056
Danas, ako bi htio koristiti tradicionalne
metode umjetne inteligencije
01:47
to solveriješiti this problemproblem,
25
95920
1216
kako bi riješio ovaj problem,
01:49
I would requirezahtevaju 10,000 --
26
97160
1456
bilo bi mi potrebno 10.000 -
01:50
I repeatponoviti, on an ordernaruči of 10,000
of these very expensiveskupo medicalmedicinski imagesslike
27
98640
4016
ponovit ću, 10.000 ovih
skupih medicinskih snimaka
01:54
first to be generatedgenerisano.
28
102680
1376
bi se moralo napraviti.
01:56
After that, I would then go
to an expertstručnjak physicianliječnik,
29
104080
2896
Nakon toga, otišao bi
kod stručnog doktora,
01:59
who would then analyzeanaliza
those imagesslike for me.
30
107000
2496
koji bi tada analizirao te snimke za mene.
02:01
And usingkoristeći those two pieceskomadi of informationinformacije,
31
109520
2096
Koristeći te dvije informacije
02:03
I can trainvoz a standardstandard deepduboko neuralneural networkmreža
or a deepduboko learningučenje networkmreža
32
111640
3656
mogao bi istrenirati duboku neuralnu
mrežu ili duboku samo-učeću mrežu
02:07
to providepružiti patient'spacijentkinje diagnosisdijagnoza.
33
115320
2136
da dijagnosticira pacijenta.
02:09
SimilarSlične to the first approachpristup,
34
117480
1736
Slično kao i u prvom pristupu,
02:11
traditionaltradicionalno artificialveštački
intelligenceinteligencija approachespristupa
35
119240
2143
tradicionalne metode umjetne inteligencije
02:13
sufferpatiti from the sameisto problemproblem.
36
121407
1449
imaju isti problem.
02:14
LargeVelika amountsiznosi of datapodaci, expertstručnjak physicianslekari
and expertstručnjak medicalmedicinski imagingObrada slike technologiestehnologije.
37
122880
4560
Velike količine podataka, stručni doktori
i stručne tehnologije snimanja.
02:20
So, can we inventizumiti more scalableskalabilno, effectiveefikasan
38
128320
4296
Možemo li izumiti učinkovitije
02:24
and more valuablevredno artificialveštački
intelligenceinteligencija architecturesarhitekte
39
132640
3296
i vrjednije metode umjetne inteligencije
02:27
to solveriješiti these very importantbitan
problemsproblemi facingsuočiti us todaydanas?
40
135960
3056
kako bi riješili ove bitne probleme
s kojim smo suočeni danas?
02:31
And this is exactlyupravo
what my groupgrupa at MITMIT MediaMedia LabLaboratorij does.
41
139040
3296
Tačno to radi moja grupa
u MIT Media Lab-u.
02:34
We have inventedizmišljen a varietyraznovrsnost
of unorthodoxneuobičajeno AIAl architecturesarhitekte
42
142360
3856
Mi smo osmislili niz
neuobičajenih UI metoda
02:38
to solveriješiti some of the mostnajviše importantbitan
challengesizazovi facingsuočiti us todaydanas
43
146240
3176
kako bi riješili neke od najvažnijih
izazova s kojim se susrećemo
02:41
in medicalmedicinski imagingObrada slike and clinicalklinički trialssuđenja.
44
149440
2200
u medicinskom snimanju
i kliničkom ispitivanju.
02:44
In the exampleprimer I shareddeli
with you todaydanas, we had two goalsciljevi.
45
152480
3056
U primjeru koji sam podijelio
s vama, imali smo dva cilja.
02:47
Our first goalcilj was to reducesmanjiti
the numberbroj of imagesslike
46
155560
2976
Prvi cilj je da se smanji
broj snimaka potrebnih
02:50
requiredpotrebno to trainvoz
artificialveštački intelligenceinteligencija algorithmsalgoritmi.
47
158560
3256
za treniranje algoritama
umjetne inteligencije.
02:53
Our secondsekunda goalcilj -- we were more ambitiousambiciozni,
48
161840
2096
Naš drugi cilj - bili smo ambiciozniji,
02:55
we wanted to reducesmanjiti the use
of expensiveskupo medicalmedicinski imagingObrada slike technologiestehnologije
49
163960
3736
željeli smo smanjiti upotrebu
skupih tehnologija snimanja
02:59
to screenekran patientspacijenti.
50
167720
1216
za snimanje pacijenata.
03:00
So how did we do it?
51
168960
1200
Pa kako smo to uspjeli?
03:02
For our first goalcilj,
52
170920
1216
Za naš prvi cilj,
03:04
insteadumjesto toga of startingpočinjati
with tensdesetke and thousandshiljade
53
172160
2056
umjesto da počnemo sa desetinama hiljada
03:06
of these very expensiveskupo medicalmedicinski imagesslike,
like traditionaltradicionalno AIAl,
54
174240
3016
ovih skupih medicinskih snimaka,
kao tradicionalna UI,
03:09
we startedzapočet with a singlesingl medicalmedicinski imageslika.
55
177280
2056
počeli smo sa samo jednom snimkom.
03:11
From this imageslika, my teamtim and I
figuredfigured out a very cleverpametan way
56
179360
3776
Iz te snimke, moj tim i ja
smo na pametan način
03:15
to extractekstrakt billionsmilijardi
of informationinformacije packetspakete.
57
183160
2736
uspjeli izvući milijarde
paketa informacija.
03:17
These informationinformacije packetspakete
includeduključeni colorsboje, pixelspiksela, geometrygeometrija
58
185920
3696
Ti paketi uključuju boje, piksele, oblike
03:21
and renderingrendering of the diseasebolesti
on the medicalmedicinski imageslika.
59
189640
2536
i predstavljaju oboljenja na tom snimku.
03:24
In a sensesmisao, we convertedPretvoren one imageslika
into billionsmilijardi of trainingobuka datapodaci pointsbodova,
60
192200
4336
Na neki način, pretvorili smo jedan snimak
u milijardu podatkovnih tačaka za trening,
03:28
massivelymasivno reducingsmanjenje the amountiznos of datapodaci
neededpotrebno for trainingobuka.
61
196560
3536
značajno smanjujući količinu
podataka potrebnih za trening.
03:32
For our secondsekunda goalcilj,
62
200120
1216
Za naš drugi cilj,
03:33
to reducesmanjiti the use of expensiveskupo medicalmedicinski
imagingObrada slike technologiestehnologije to screenekran patientspacijenti,
63
201360
3856
smanjiti upotrebu skupih tehnologija
snimanja kod pacijenata,
03:37
we startedzapočet with a standardstandard,
whitebela lightsvetlo photographfotografija,
64
205240
2856
počeli smo sa standardnom,
fotografijom sa bijelim svijetlom,
03:40
acquiredstečena eitherbilo to from a DSLRDSLR camerakamera
or a mobilemobilni phonetelefon, for the patientpacijent.
65
208120
4336
napravljenom profesionalnom kamerom
ili mobilnim uređajem, za pacijenta.
03:44
Then rememberzapamtite those
billionsmilijardi of informationinformacije packetspakete?
66
212480
2456
Sjećate li se onih
milijardi paketa informacija?
03:46
We overlaidpreležao those from
the medicalmedicinski imageslika ontona this imageslika,
67
214960
3536
Spojili smo ih iz medicinskog snimka
sa ovom slikom,
03:50
creatingstvaranje something
that we call a compositekompozitni imageslika.
68
218520
2520
i napravili nešto što zovemo
kompozitna slika.
03:53
Much to our surpriseiznenađenje,
we only requiredpotrebno 50 --
69
221480
3296
Na naše iznenađenje,
bilo je potrebno samo 50 -
03:56
I repeatponoviti, only 50 --
70
224800
1336
ponovit ću, samo 50 -
03:58
of these compositekompozitni imagesslike to trainvoz
our algorithmsalgoritmi to highvisoka efficienciesefikasnosti.
71
226160
3840
ovih kompozitnih slika kako
bi algoritme učinili jako efikasnim.
04:02
To summarizesumirati our approachpristup,
72
230680
1336
Da rezimiramo naš pristup,
04:04
insteadumjesto toga of usingkoristeći 10,000
very expensiveskupo medicalmedicinski imagesslike,
73
232040
3176
umjesto da koristimo 10.000
vrlo skupih medicinskih snimaka,
04:07
we can now trainvoz the AIAl algorithmsalgoritmi
in an unorthodoxneuobičajeno way,
74
235240
3016
možemo istrenirati algoritme UI,
na jedan neobičan način,
04:10
usingkoristeći only 50 of these high-resolutionvisoka rezolucija,
but standardstandard photographsfotografije,
75
238280
4256
koristeći samo 50 ovih slika
visoke rezolucije,
04:14
acquiredstečena from DSLRDSLR cameraskamere
and mobilemobilni phonestelefone,
76
242560
2496
snimljenih profesionalnom
kamerom i mobitelom,
04:17
and providepružiti diagnosisdijagnoza.
77
245080
1536
i postaviti dijagnozu.
04:18
More importantlyvažno,
78
246640
1216
Još bitnije,
04:19
our algorithmsalgoritmi can acceptprihvatiti,
in the futurebudućnost and even right now,
79
247880
3136
naši algoritmi mogu prihvatiti,
u budućnosti ali i sada,
04:23
some very simplejednostavno, whitebela lightsvetlo
photographsfotografije from the patientpacijent,
80
251040
2816
jednostavne fotografije sa bijelim
svijetlom od pacijenta,
04:25
insteadumjesto toga of expensiveskupo
medicalmedicinski imagingObrada slike technologiestehnologije.
81
253880
2440
umjesto koristiti skupe
tehnologije snimanja.
04:29
I believe that we are poisedstaložen
to enterunesite an eraera
82
257120
3096
Vjerujem da postepeno ulazimo u vrijeme
04:32
where artificialveštački intelligenceinteligencija
83
260240
1936
gdje umjetna inteligencija
04:34
is going to make an incrediblenevjerovatno
impactuticaj on our futurebudućnost.
84
262200
2536
predstavlja veliki značaj
za našu budućnost.
04:36
And I think that thinkingrazmišljanje
about traditionaltradicionalno AIAl,
85
264760
2456
I mislim da uz razmišljanje
o tradicionalnoj UI,
04:39
whichšto is data-richbogate podatke but application-poorprijava-siromašna,
86
267240
2776
koja je bogata informacijama,
ali siromašna primjenama,
04:42
we should alsotakođe continueNastavite thinkingrazmišljanje
87
270040
1536
trebamo nastaviti razmišljati
04:43
about unorthodoxneuobičajeno artificialveštački
intelligenceinteligencija architecturesarhitekte,
88
271600
3016
o neobičnim metodama
umjetne inteligencije,
04:46
whichšto can acceptprihvatiti smallmali amountsiznosi of datapodaci
89
274640
1936
koje koriste mali skup informacija
04:48
and solveriješiti some of the mostnajviše importantbitan
problemsproblemi facingsuočiti us todaydanas,
90
276600
2936
i rješavaju neke od najvažnijih
problema današnjice,
04:51
especiallyposebno in healthzdravlje carenegu.
91
279560
1256
posebno u zdravstvu.
04:52
Thank you very much.
92
280840
1216
Mnogo vam hvala.
04:54
(ApplausePljesak)
93
282080
3840
(Aplauz)
Translated by Mario Filipović
Reviewed by Ivana Korom

▲Back to top

ABOUT THE SPEAKER
Pratik Shah - Medical technologist
Dr. Pratik Shah creates novel intersections between engineering, medical imaging, machine learning and medicine.

Why you should listen

Dr. Shah's research program at the MIT Media Lab develops scalable and low-cost diagnostics and therapeutics. His ongoing research areas at MIT include: 1) artificial intelligence and machine learning methods for detection of cancer biomarkers using standard photographs vs. expensive medical images; 2) unorthodox artificial intelligence and machine learning algorithms to design optimal and faster clinical trials and to reduce adverse effects on patients; and 3) low-cost and open source imaging devices, paper diagnostics, algorithms and mobile phones to improve public health and generate real-world data.

Clinical studies with Pratik's medical technologies have revealed "missing sick" patients, who otherwise remain undiagnosed in conventional healthcare settings. Dr. Shah's graduate and postdoctoral research contributed to the discovery of a vaccine component to prevent pneumococcal (Streptococcus pneumoniae) diseases; the identification of new pathways, technologies and metabolites as antimicrobials to target gastrointestinal infections; and a nonprofit to deploy a low-cost water quality test for the developing world.

Past recognition for Dr. Shah includes the American Society for Microbiology's Raymond W. Sarber national award, the Harvard Medical School and Massachusetts General Hospitals ECOR Fund for Medical Discovery postdoctoral fellowship, the AAAS-Lemelson Invention Ambassador Award and a TED Fellowship. Pratik has been an invited discussion leader at Gordon Research Seminars; a speaker at Cold Spring Harbor Laboratories, Gordon Research Conferences and IEEE bioengineering conferences; and a peer reviewer for leading scientific publications and funding agencies. Pratik has a BS, MS, and a PhD in Microbiology and completed fellowship training at The Broad Institute of MIT and Harvard, Massachusetts General Hospital and Harvard Medical School.

More profile about the speaker
Pratik Shah | Speaker | TED.com

Data provided by TED.

This site was created in May 2015 and the last update was on January 12, 2020. It will no longer be updated.

We are currently creating a new site called "eng.lish.video" and would be grateful if you could access it.

If you have any questions or suggestions, please feel free to write comments in your language on the contact form.

Privacy Policy

Developer's Blog

Buy Me A Coffee