Page 27 - Aesthetics&Dermatology_11_2018

RESEARCH REPORTS
26
Papier told MobiHealthNews. “So
we never see the image because
of confidentiality. So I think on the
consumer side there’s an opt-in
where the user can click a box
where they say they’re willing
to train data and can create a
feedback loop and start training the
system. We’ll be getting going with
that this year.”
Cleaning up messy images
What most researchers and
developers want is an AI that
can look at pictures taken with a
smartphone in a home environment.
And smartphone camera pictures
can look very different from one
another, even if they’re of the same
thing. Differences in angle, lighting
and focus need to be accounted for. cross-validating what we’re doing in
The Mars rover ... if you think
about what it’s doing, it’s actually
imaging the planet surface using
multiple cameras and creating a
reconstruction,” Carlo Perez, CEO
of AI wound care company Swift
Medical, said. “But if you think
about the surface of Mars and you
put it on my body, it actually looks
like a wound. Except that a wound
is far more complex to interpret than
the surface of Mars. It’s actually
a harder problem because you
have to account for variations in
skin color, different types of tissue,
different levels of moisture. It’s a
very hard vision problem that can
be augmented and solved with AI.”
For dermatology, Sun and the team
at First Derm have been working
on ways of computer-treating the
image to improve accuracy.
“There’s a couple things we can
do,” she said. “First, there’s general
lighting guidance that we provide
users. But also, generally, people
take a larger picture than you
need, like they’re not just going to
take a picture of the boil. So you
can zero it against the best base
reference that you have which
is another portion of their skin in
the photo. And there’s a couple of world,” Papier said. “Those rashes
imaging methods that we have been look the same, you have a patient
experimenting with.” who has a viral infection from a
rare mosquito-borne disease in the
Dermatologists today are especially Caribbean here and you have one
aware of the risk of creating who has a drug reaction to penicillin,
algorithms that only work with the rashes look near identical. So
certain races of people, an easy the machine learning can put you in
and dangerous training bias to the neighborhood, but you need the
introduce. history.”
“It’s something that we’re trying to Ultimately, what researchers are
be really cognizant of,” Sun said. hoping to create is a largely accurate
“This is something that gets left out algorithm that can help give piece
of the AI conversation sometimes. of mind to people concerned about
So obviously we’re not going to their skin — and possibly too
pretend we can classify someone’s embarrassed to seek other help.
race or cultural background from
an online photo, but we are doing “The end goal is first to develop a
our best to make sure that we are pretty robust image classifier,” Sun
using concepts like the reference said. “So there is an accessibility
zeroing — making sure we’re not point with this. Especially with
using one standard as the standard something like pictures of your
for everything and also kind of genitals, it can be an embarrassing
topic, or it can be something you put
different populations, so it’s not just off seeing a doctor for or you don’t
‘We’re going to try to average out really share with anybody or you just
these effects.’” hope for it to go away. … So we’re
hoping to build something that you
“One of the things we’re really proud don’t have to pay tons of money for,
of at VisualDx is for the last 18 that will give you a pretty good idea
years we’ve been cataloguing the of what’s going on and that can be
spectrum of disease in people of all a kind of intermediary to see if you
colors,” Papier said. “So the goal need to pursue more aggressive
of Aysa is to give a really unique treatment. So we’re targeting
experience to someone, if they have accuracy goals of around 80 percent
light skin or dark skin. We give them for that, across a couple common
a very custom experience.” categories of dermatological issues.”
The limits of AI accuracy While there is a risk of giving false
negatives that could endanger
Even once a computer can make patients, there’s also no alternative.
an apples to apples comparison Patients who are concerned about
between different lesions in different their health, but not concerned
lightings taken with different enough to go to a doctor, are going
smartphone cameras, most to look somewhere.
researchers don’t think it will be
able to make a perfect diagnosis. “The fact is that most people, when
they or a family member is sick,
One reason for that is that even or in the case of dermatology they
human dermatologists disagree have a rash, they go to Google,”
sometimes, especially in the Papier said. “They look right away
absence of additional information. on Saturday or Sunday and the
question is can we give them
“There are circumstances where reliable, helpful information that is
the rash looks identical and the safe for them that’s better than what
machine learning will not know the they’re currently doing?”
difference between a rash from a
drug reaction or a rash from a viral
infection from traveling around the
27