Player is loading...

Embed

Embed code

Transcriptions

Note: this content has been automatically generated.
00:00:00
i mean i i think seventy minute wins and um to remarks up phone device stared you
00:00:08
love that and we do this because i lost my glasses on the way here
00:00:11
at and the second reason of course is i'm german norse agenda so does my happy face so i have nothing against you
00:00:19
that's just why yeah i always like to start i understand you have a ah number
00:00:24
for interesting talks only behind you and i want to start a small exercise
00:00:28
um that helps you to me dude your brain a little bit if it's called
00:00:31
mind wondering it requires that you put your phone just away for a second
00:00:35
get up straight in your shared maybe go with the shallow that back and make eye contact with me for it
00:00:42
wait for translation but i can see you all please take justin seconds to make eye contact you sit up straight that's very important
00:00:51
make eye contact with me
00:00:53
and we all closer nice
00:00:56
and you think of the person that is most important and most other than to you in your life
00:01:04
great image in your head of the person
00:01:07
maybe you wanna smile maybe not
00:01:12
and you open your eyes again keep that image
00:01:15
if i close my nice i seem under front of um a good luck for me no
00:01:23
and i live very very easy simple principle my life happy wife happy life that's pretty
00:01:31
hard to go buy a if you are from the syphon from pain every day
00:01:35
because when a matter five years ago should last actually since you last fourteenth
00:01:40
and as a good portion that was going for the doctors and and
00:01:45
it took care five years twenty two visits to the doctor for
00:01:49
to come finally to the white your nose is in the treatment between as was done for easily was actually and the mitre wasn't
00:01:56
and combination dropped was intolerance so almost with and and if you would like to say so
00:02:02
but however people overlooked wasn't taken seriously and was not treated the white whale
00:02:09
and my point is if you don't know what you have a
00:02:12
few diagnose is not on might know treatment can start whatsoever
00:02:17
so we believe getting the treatment getting diagnosis right is the most important part start in the white house journey
00:02:25
and we believe that is a problem for number of reasons and one no one
00:02:30
can really hard hard to argue if it's just the sheer number of complexity
00:02:35
so we have tons of symptoms thousands of known disease
00:02:41
and just the simple fact that a normal doctor can maybe have one thousand one thousand five hundred of them in his head
00:02:49
and no one is white men would make the point is you can actually we oversee all kind of red it's nice
00:02:55
and if you consider that complexity always make an um example on one show 'cause it just a member everyone of your image your hat
00:03:03
you would print it out as a possible you know like this this kind of presence you get for your spouse when
00:03:07
you have not that idea and we all in this room which take is possible and turn it to the floor
00:03:13
and then ask someone to put it all together back again because just sometimes symptoms fitting to different disease but you
00:03:20
don't know what you're looking for up to the top of going to deposit light is actually very challenging
00:03:27
we believe that technology can help with that because it helps the doctor to
00:03:33
oversee the complexity and allows him to come to light it says
00:03:38
there's emission of later we believe it is a fundamental right for every person on this planet to get to my joke noses
00:03:45
and then just started like different and we believe we're here to support patients as well as doctors
00:03:56
how we view it
00:03:58
we do that we and application does called a got checked if it's in a i based chat part
00:04:05
that allows you to gather information into a series of questions lead you to differential
00:04:10
diagnosis that tells you what might be the best fits forces into thing
00:04:17
there's a second product should call a uh for doctors which allows exactly the same uh engine but with a different
00:04:24
approach because if you're patient they wanna take being hand by hand because it's the subject if you're not real
00:04:30
you don't care particularly much about disease definitions and symptoms
00:04:35
but if your doctor you don't need to be taken by the hand actually see patients walking to the door and have already
00:04:40
pretty good first imaging maybe you just the one or two flops uh to be sure to you deserve a user flow
00:04:48
but the interesting part is not how the product is journal part are to the people interesting parliament focus on today's how it works
00:04:55
because they i think they dumb machine learning all passwords nowadays that have been following
00:05:01
aunt conferences we started working on this actually back in two thousand eleven
00:05:06
we started to fifty doctors netting out every known centre
00:05:11
make sure that every known disease and adding probabilities to read based on the
00:05:16
um medical research and the shadows up there and constantly updating those
00:05:21
based on that we visit developed a neural network that is then put forward fifty patients input data
00:05:27
to it and then be very fight it still have medical experts in the in the background
00:05:33
so we strongly believe didn't approach entirely focusing on data
00:05:37
algorithms would never work and the little good results
00:05:40
again a simple picture to illustrate that um if you just take data and generalised experience
00:05:47
you might end up with some general rules but as what have we have done in the middle age you know generalising experience
00:05:53
we believe you need a very fine experiment object the full truth in order to
00:05:58
make progress and uh chief the number one level comes you want to have
00:06:05
so you have to strangulation of approach of the medical experts you have the data and then you combine of that but
00:06:13
i mean not go into the details of that but explaining a little bit where we start so
00:06:20
you have the possibility nader to get symptoms like direct information from the patient all the doctor
00:06:26
but could also start early and get on health information jean data yeah data
00:06:31
to be you know change the probabilities the likelihood for certain patients how certain sentinel disease
00:06:38
an easy example someone uses ada in africa malaria are is a much more
00:06:43
likely case and if you use it for the same sentence in europe
00:06:49
what then happens is not the pasta tendency of questions so a
00:06:52
that is not fixed and focus on pushes you certain direction
00:06:56
what we do is we asking you questions in order to minimise the possibilities
00:07:01
rooms for diagnosis that means of every question eliminating thousands of options
00:07:07
and because we do so we are able does not very important to show
00:07:11
you transparently in the and why we have come to certain result
00:07:16
because of course nobody's forever also i wasn't so i'll use needs to
00:07:21
be able to be transparent unexplainable to its doctors and patients
00:07:25
and shoulder way of reasoning to spot mistakes and also maybe um for example if the pay she you know
00:07:32
uses along information doctor should be able to go back and say that this information needs to be changed
00:07:38
what then happens is become one of the one too fancy have despotic regimes we come
00:07:43
up with a differential diagnosis that tells you what is the most likely fit
00:07:48
to your conditions and what is the most um probable
00:07:54
that means arises for example is not very common so it's not very likely that you have it
00:07:59
but very often it's the best fit to explain certain um patient um
00:08:05
case that's also the reason why the most often overlooked because if you have to treat hundred patients but day returning
00:08:12
to what what is most common and what you see normally every day and most of the cases your light
00:08:21
if you look real deep it is unfortunate is print is not very big
00:08:25
a little bit here is is in the end how information accessed
00:08:28
mean reprocessing thousand tolls of of data points and information and
00:08:32
in the end we illustrated in a very easy understand the car for the show what are the most one sentence
00:08:39
to which condition this speaking to and you see the figurine like like what is
00:08:43
like very fine a certain properties and what is falsifying sort of budgets
00:08:48
and if that's simple trick let's say we are able to choose huge number
00:08:53
of complexity and make it accessible slow for patients also just for
00:08:59
we didn't them off validation studies pilots and its appearance of that don't want to show you one particular one today
00:09:06
some of the medical university of an offer and the idea was what
00:09:11
would happen if we choose doctorates from the last ten years
00:09:15
and gives the same information the doctor has written down to a data and compare the
00:09:19
results to the actual outcome and of course uh you would suspect it wouldn't be
00:09:25
here on this conference it was not a success allies the most dramatic case we found
00:09:32
was mister k. is the case for symptoms first started in nineteen ninety five
00:09:38
starting getting treatment in two thousand two thousand one and was diagnosed in two thousand fifty
00:09:46
up to see very very long painful john doing discourse um he lost his job about depression
00:09:52
um i got divorced and had actually very visible um life and uh uh you explain
00:09:59
it the most middle part for him was not knowing what was going on
00:10:04
after being widely diagnosed however treatment started life signifying proved
00:10:10
i don't for this case we found that a that would be able to find the
00:10:15
right there course like you did before related disease already in two thousand one
00:10:23
i don't claim that that would necessarily automatically to the white feet in in two thousand one
00:10:28
but would would have been a very good friend for the treating doctors that if
00:10:31
the you know the first iteration doesn't work maybe to look at this direction
00:10:35
on average we find that around seven years of stifling on for finding
00:10:40
and to find a lighter those could be saved a using technology
00:10:45
actually i we won't doesn't out from university cleaning often over to five more years clippings
00:10:53
who skip that part ada has been compared to of course the competitors validation
00:10:58
studies will most people found over and over again is that our curse
00:11:02
a little much higher and the covering much much more disease than our
00:11:06
competitors i think the main reason here's why you is our history
00:11:11
course it was founded coming out of the max planck institutes and how many news foundation
00:11:16
so uh background is in signs not necessarily business so i was not
00:11:20
set up to you know make a fortune or make fast except
00:11:26
our phones that reason but i i do your that way of disease are
00:11:30
where but if you think of it is that ten percent always combined
00:11:35
you know make up a a tort reform two hundred million of all the dust silently suffering
00:11:40
among us he what's considered this is like sub saharan africa that's the minimum wage
00:11:44
just because you could con put them all in one place and make it visible how much is people suffering
00:11:49
uh not enough money is that cater to research and treatment and support for these people
00:11:59
but of course the technology has no use it's not used so we looked into case of energy hasn't worn out
00:12:07
in u. k. for example the number of clicks and show like two patients actually want to use that
00:12:12
i found yes even elderly people will say okay stare approach technology not different from young people are we
00:12:19
found that the oldest i'm actually at an shouldn't eighty eight is exactly but doctors next important question
00:12:26
we found it is and get 'em numbers positive feedbacks from doctors
00:12:30
using damn windy because all claims not that we want to
00:12:33
replace the doctor our claim is that we hope him very difficult
00:12:37
unchanged ration um to help us patient the test extent
00:12:44
where he was standing right now few numbers so we have around approximately so
00:12:50
to just thirty five thousand patients coming in every day so we have
00:12:55
i think i always updating dislike but never get a to this day
00:12:59
the scroll thing right now for five point four million assessments
00:13:03
and in there to see the number of instruments is lower that means we have a retention week of
00:13:07
around thirty percent at the moment so people actually use that i find it useful and become
00:13:12
we have around ninety thousand positive uh feedbacks and um relation that patient is also how we
00:13:19
is a very important channel loss of getting feedback people very often um tell us i went to
00:13:25
ada at the writer president adopt and he actually said it was exactly what data suggest that
00:13:33
does that just number of examples from people were getting
00:13:37
i'm always make this show control if you have a bad day just go to the user feedback
00:13:41
and we to it uh that's very motivating um but of course
00:13:46
it's at the moment um uh because all anecdotal evidence
00:13:52
so we don't have yet i'm a systematic approach of tracking patients to complete have system
00:13:58
we working two words that are popping up of health insurance and fifth operations
00:14:04
so we just signed to speak for example of the largest health insurance in germany plastic or can cause a lot of women
00:14:10
that would allow us channel to get the feedback automatically folder i'm
00:14:15
sure people back was on ten point four million people
00:14:19
do you ski is a very different so it could be used um to improve the quality of care
00:14:27
it could be a great tool of linking also solution is to improve the um cheap enough
00:14:33
clear the future of pull tractor or if the disease management especially for chronic disease
00:14:38
and could be of a great way of finding patients to do by choice at the time
00:14:44
we basically see it as an i'm substitute for just getting your symptoms and wanting to the
00:14:49
emergency uh because you think you have you know can send your time next day
00:14:57
for the doctors um of course the most important decision support
00:15:03
um but we also believe it could help to use on the
00:15:07
steaks and open up having of cases button colour put together
00:15:12
and of course data could be also very useful part from abusing other
00:15:15
technologies so we know jehovah's advancing very fast speed up it's
00:15:21
okay and um do the question was how do you have that complexity and how to you as a doctor
00:15:28
um we don't have time to dive into everything the solution how do you decide which to work with machine part of
00:15:37
we wanna number of prices i was forced to have the slide in here
00:15:40
i don't think it's perfectly useful um more interesting is maybe where the joining going to words to
00:15:45
and and just previously the hurt already of the power of including
00:15:49
genetic profiling um work on that as well because we think
00:15:54
the possibilities there the question of how we're gonna use it in every day practised and we believe it's gonna be very hard
00:16:00
uh for doctors to makes it even if you would be finally make
00:16:03
the stuff of making patient information available at all times and everywhere
00:16:07
how do you assess that if you have an average seven eleven minutes to do for patient if you're jennifer just
00:16:14
we could help with that because all probability models would allow that every single bit of information are getting so
00:16:19
loud change the probabilities and then you can very easily show you what to do with that information
00:16:26
no no we mention it will be very passionate about is um building
00:16:31
alliances um to how to detect and map out where disease
00:16:36
we all to make you believe that the trend to what's personalised medicine what actually
00:16:40
turn almost every disease into where disease if you like to say so
00:16:44
also just very nice example we all different from the outside most likely all different from the inside
00:16:50
and at the moment we half of course to time resource limitation we have a one size fits all approach
00:16:56
and we think this was can change if it's gonna change also the treatment is gonna
00:17:00
change and so where diseases in the moment the end of next for topic
00:17:04
but the thing with the things we gonna learn there again ultimately benefit of us
00:17:10
thank you very much for your attention
00:17:18
and i apologise up front icon speak french i have french different once but it may be worth
00:17:23
mentioning crucial problem you can ask enjoy can get there yet transit with so many question
00:17:30
so we would stop on the left side and then the right i thank you very much for presentation i was
00:17:38
just wondering what is the regular three stages of the
00:17:40
project of your or your medical devices diagnostic tool
00:17:45
oh you see by regulators basically that's um that's a good question we
00:17:51
certified as a medical product um be to try to um to
00:17:57
x. think very carefully what we do that this is not an substitute for going to the doctor speech is a suggestion
00:18:04
in most in most like cases take this and go to the doctor and
00:18:07
get proper care but engine you have a very special so every v.
00:18:11
we're at the moment in u. k. in germany and us and have
00:18:15
number projects wanting in tanzania kenya and look into our market so
00:18:19
i'm in germany to situation that or it was all base is quite different because
00:18:23
we have the whole from distant now once we chancellor you heard of that
00:18:27
and so we have suddenly a situation where it feels like everything goes
00:18:31
so to form insurances to hospitals doctors everyone feels like we have waited ten
00:18:37
years we've debated a long time now is the moment to move forward
00:18:40
where you normally would feel traditionally you will fit a lot
00:18:43
of resentment and be very have very for careful steps
00:18:47
now and you we feel that right now it seems to be the momentum everyone wants them for that
00:18:54
thank you lot thank you for your your talk so in terms of actual objective
00:19:00
performance can you share some numbers on how well you were your machine operates
00:19:06
um you need a a lot of accuracy light accuracy your
00:19:09
information or anything like that yeah um it differs differs
00:19:14
depending on the disease on the condition and the environment numbers and the quality of information getting
00:19:20
we have performed around seven validation studies and the low was strange was thing seventy six percent
00:19:27
and then the best cases was up two hundred percent you know depends on if you wanna detect or
00:19:32
fever or you want to detect uh uh caps um
00:19:38
i also to think it's would be um too much simplicity to say we have to soccer sort of
00:19:43
of ninety percent and it's just it's just not true depends on the which discipline you look at
00:19:48
and it depends on the quality of information we have another ten politicians something's coming up
00:19:54
uh to look at different fields because we think it's a very good way to learn and
00:19:59
if we often often get confronted what happens if you do a mistake you
00:20:04
know you don't have hundred percent so obviously also ah sometimes wong
00:20:07
and i have to openly admit where a uh house i'm still to the
00:20:12
works especially when they have come abilities so when you have you know
00:20:16
several disease interacting with the java because that's very hard to model and a certain extent
00:20:22
but the reality is also doctors not hundred percent and as long as we're working in partnership another displacement
00:20:28
we feel we adding much more well you may causing harm also we didn't have a single cases someone was
00:20:34
mistreated and you know something that happened he was not treated uh in a certain extent so we'll
00:20:41
i think hope that answers your question if anyone's interested i'm happy to send on after
00:20:45
it's more slides on the validation part as well okay thank good next question
00:20:52
used to go to a question techniques we've talked with the first one is the no if
00:20:57
the uh you're probably double will probably do tricks is it available for p. beatrix
00:21:04
um dashboard is freely available for everyone and it's used with
00:21:09
the doctor version it's not willing to sit and um
00:21:13
um i feel that i think paychecks behalf project going on in the nick at the moment
00:21:19
but there we are not in the rolling out face yes the the pardon face
00:21:23
not because the product is not ready but because the infrastructure is not ready
00:21:27
you know because if you put it in a in a in in day practised and you just know where to
00:21:31
put in the data and you know do something with little reported back um that's gonna be difficult um yeah
00:21:38
so yes but in that it okay the next question who who who profess stuff e. n. c. i.
00:21:50
we're still working on so we know that challenges that you have a new one be probably be
00:21:57
the beat and a lot of hype but because we lack date now where we
00:22:02
can make sure yeah q. i. e. like rachel and you get ready
00:22:07
so we have an x. that upbeat unfortunately i review of their biggie like we could name
00:22:16
oh the second show the like poor performance for now
00:22:23
probably that we may come when i but i see now i made sure why
00:22:30
let me think about now in full of not
00:22:36
i thought we should have yeah you put o. e. cool
00:22:46
there's no question i'm most afraid of because i know we have a pretty good answer for that
00:22:50
but i'm just not the right person and you see a a eight iverson very interesting
00:22:55
antidepressant gee i have people coming from the data field machine learning but also medicine
00:23:01
and for example this particular problem or head of research would be perfect johns it um but i would have to say
00:23:07
please send an email and i will fall off of that on this in that i would try a simple and
00:23:12
this is the answer is um if we put in objectively data like lab tests as well and so we not
00:23:18
we line alternately only on experts and we half by by
00:23:21
now five point four million cases to verify against
00:23:25
thing we have stocking with most just that the madison treatment
00:23:28
and also um possible turkeys is also culturally present so
00:23:32
if you look at fox on the tropical disease which we do now enough uh current and then yeah especially
00:23:36
um you can't have the same probabilities and the same suggestions like you have in your
00:23:42
so for example this a star of working refill for positions and yeah
00:23:46
to prove that the situation and also the corporation the flat just provide us
00:23:52
thinking and i wrote me and also tanzania this will take time we're way of the challenges and if you wanna
00:23:58
go deeper i invite you to write an email then happy to link you had of research thank you okay
00:24:06
you should really try to have both the big videos you put the which shows report we the with the
00:24:11
tomato collected goodwill much the both of those limits you know thank you were which will be the fish

Share this talk: 


Conference program

Mots de bienvenue
Sébastien Mabillard, CEO | Swiss Digital Health
15 June 2018 · 9:07 a.m.
Mots de bienvenue
Victor Fournier, Chef de Service de la santé publique | Canton du Valais
15 June 2018 · 9:11 a.m.
Mots de bienvenue
Jean-Albert Ferrez, Président | Fondation The Ark
15 June 2018 · 9:19 a.m.
Mots de bienvenue
Laurent Sciboz, directeur Institut Informatique de Gestion | HES-SO Valais/Wallis
15 June 2018 · 9:24 a.m.
L'écosystème d'innovation ouverte du CHU Sainte-Justine: une grande communauté gagnante!
Kathy Malas, resp. de la Plateforme de l’innovation et des Fonctions des maladies chroniques et aiguës | CHU SAINTE-JUSTINE (Canada)
15 June 2018 · 9:33 a.m.
Les livraisons par drones : vers une amélioration de la logistique dans le domaine médical
Janick Mischler, Program Manager | LA POSTE SUISSE
15 June 2018 · 10:07 a.m.
E-health et intégration des soins
Marc Cikes, CEO | MEDBASE ROMANDIE (Suisse)
15 June 2018 · 10:32 a.m.
Questions réponses
Remi Gauchoux, Business Development Director - Carenity
15 June 2018 · 11:46 a.m.
Futur de la santé mobile
Dr. med. Patricia Sigam, CEO & Co-founder, digital Med-Lab
15 June 2018 · 11:54 a.m.
Democratizing Data-Driven Medicine
Tarik Dlala, VP Marketing, Sophia Genetics
15 June 2018 · 12:14 p.m.
Ada inside
Vincent Zimmer, Ada Health, Berlin
15 June 2018 · 12:39 p.m.
biospectal, the optical revolution in hypertension monitoring
Prof. Patrick Schoettker, CMO. Biospectal
15 June 2018 · 2:06 p.m.
3D Printed Medicines: A Digital Pharmacy Era
Sarah Trenfield, MPharm, Senior Formulation Scientist, FabRx Ltd.
15 June 2018 · 2:23 p.m.
The Digipharm experience
Ahmed Abdullah, CEO & co-founder, Digipharm, Basel
15 June 2018 · 2:45 p.m.
Enabling the rehabilitative revolution
Dr. Manfredo Atzori, HES-SO Valais Wallis
15 June 2018 · 3 p.m.
Team Gamified Multi-sensory Stroke Rehab
Jean-Luc Turlan et J-P. Ghobril, Lauréats Arkathon 2018
15 June 2018 · 3:22 p.m.
Secure and Trustable EMR Sharing using Blockchain: Open Challenges and Lessons Learned
Alevtina Dubovitskaya, HES-SO Valais-Wallis
15 June 2018 · 3:38 p.m.
Conclusions
Sébastien Mabillard, CEO | Swiss Digital Health
15 June 2018 · 3:59 p.m.

Recommended talks

Les réseaux de soins de patients: vers une optimisation des modèles de soins
Evert Jan Hoijtink , CEO de Portavita.nl
7 June 2013 · 3:11 p.m.
Health care is more than medical care!
Prof. Dr. Sebastian Probst, DClinPrac, RN, Professor of tissue viability and wound care at the University of Applied Sciences Western Switzerland, Geneva, Switzerland
5 April 2017 · 2:07 p.m.