[Music]
hi thanks very much for coming I’m gonna
tell you a little bit about one thing
that’s next for science and I’m gonna
use an example from my own research but
what I’m talking about is broadly
applicable to science writ large I’m
gonna talk to you about very big data
that’s been set free data for my whole
career has been big I’ve always had full
diskettes now discs now raids when I
started they were 92 kilobytes and I
filled them then a hundred megabytes
then 10 gigabytes and now huge amounts
so as physiologists we looked at data in
the kilobytes to may be megabytes worth
of data genomics which has done a
fabulous job both of being big and of
sharing and being open works in the
megabytes thinking a bytes range I’m
interested in shape I’m interested in
the shape of skeletons and the soft
tissue that surrounds them and those
data are huge so they run from the
gigabytes to terabytes
for single datasets and date has always
been big relative to our capabilities
but until very very recently data was
not free data was locked into little
silos in scientists labs
where you couldn’t get at them where I
couldn’t get at them where their
colleagues couldn’t get at them where
their spouses and children couldn’t get
at them nobody could get at these data
sometimes ten years later they couldn’t
get at so I’m gonna talk to you about
big morphological data sets the data set
I’m gonna talk about is computed
tomography scans CT scans first thing I
tell you what a CT scan is which
requires some fabulous finger puppetry
behold a photorealistic bird look only
at the screen and you see the bird and
it is mysterious to you because you
didn’t see me come in with a bird but
now there’s one but if I freeze my hand
in place now look only at the screen and
rotate it through a hundred and eighty
degrees watching just those shadows in
your brain you can now not only see that
that wasn’t a bird you can put your
fingers in the same position as my
fingers were in you can deconvolve those
shadows into the 3d shape of my hand CT
scanning is exactly that except instead
of light it uses x-rays so instead of a
black and white shadow there’s a
grayscale shadow with the darkest bits
being the most radio-opaque and the
lightest bits being the least
radio-opaque
so the CT scan generates a 3d volume of
density
the end result looks like that that is a
piranha a beautiful fish that I scanned
because I wanted to see its teeth which
are the business end of the thing right
that’s what everyone wants to see
but the moment I looked at that scan I
realized I knew something about how it
lives way over on the side there you can
see two ribs that have been bisected
because this is a schooling fish that’s
really well armed and when they have
their little discussions amongst each
other they lead to serious wounds so
these models of skeletons can tell us
how fish lives and I’m interested in CT
scanning all of the fishes not every
individual but every species its 33,000
species of fish and at conventional CT
scanning rates we’re talking about a 45
year project so just about when I hit
middle age I should be done I plan to
take up exercise between now and then
the killer is this effort should cost
between 3 and 300 million dollars and as
much as I deeply love fish and I really
do I mean I’m deeply fascinated with
them I couldn’t justify spending those
kinds of funds on scanning fish
skeletons so we got a choice we can
either scan fewer species or we can scan
faster and I am fundamentally not a
fewer kind of person
so we raised some philanthropic funds
actually a very large set of donations
allowed us to buy a micro CT scanner
that lives just down the hall in my
office and from this scanner I
discovered two truths to what should
have been obvious truths but which to me
at the time
we’re not obvious at all the first of
these is that all CT scanning
interrogate s– a cylindrical volume how
many people here have actually been in a
CT scanner just give a raise van oh wow
huge numbers that’s awesome so you’ve
laid there you lay there in a bed and
either the bed goes into a doughnut or
in some models the donut runs down the
bed that produces a beautiful
three-dimensional image of the offending
part of your body what you may not know
is that everything inside the donut
everything was scanned you and your
offending part the bed the air around
all of it was scanned but we know what
air looks like and we know what beds
look like so this was terribly
inefficient we really should have piled
at least half a dozen people on there
with you run them all through it once
because everything was getting scanned
so that was the first realization the
second realization was more subtle and
it it turns out to be true that you
don’t want all of the resolution anytime
you take a picture there’s a certain
size for those pixels in CT the an
analogy is voxels that is volume
elements and all of you know when you
get a new camera what do people ask you
oh how many megapixels is that all right
that’s what they ask the first question
they don’t ask like how much was it do
you like it that you take an icepick
you know how many megapixels there are
another resolution if you own a CT
scanner that’s the first thing there so
what’s the resolution second thing they
ask is how big a thing can I put in
there as though there were some universe
where they would want to get five
thousand slices per millimeter of an
entire dead hippopotamus which would be
more data than fit on all the hard disks
in the United States right now you don’t
want all the resolution so how much do
you want I mean that’s a good question
one that we’ve said about to answer we
looked at a lot of studies and we found
that it depends totally on what you’re
doing with it and so what we had to do
was figure out what the worst case was
what what’s the most resolution you
could want and it turns out that the
most resolution you want for practical
purposes is about two hundred and fifty
slices across the head so give me a fish
I can measure the width of its head and
I can tell you what resolution you want
to scan it that means that I can pre
print cylinders that I can pack with
fish the cylinder on your right is a 30
micron cylinder 33 slices per millimeter
on your left a cylinder that’s larger
it’s a 50 micron cylinder it gives you
20 slices per millimeter and I can just
pack the fish in there this doesn’t work
with humans in healthcare it’s an
innovation I’ve proposed but it turns
out that people object dead fishes don’t
it’s one of the major reasons I work
with them
the resolution you aim for it depends on
the fish size which means when you bring
hundreds of fish to my lab to scan them
the first thing I have you do is arrange
them in size order we then pick the
specimens that are going to be scanned
at a particular resolution and we start
loading up the cylinder we can fit 10 to
20 specimens in that cylinder and when
we scan it we get this
we have reimagined that staple of the
mid week the fish burrito that is
somewhere around 20 individual species
in a group of sculpins scanned by a
graduate student named Thaddeus boo sir
Thaddeus came to the lab and and and we
developed this technique this is the
very first fish burrito that was ever
scanned I know this because it is
missing a key innovation that came about
with the second fish burrito that was
ever scanned little radio-opaque labels
so that we could tell one fish from
another because it turns out that only
fattiest could do this next trick using
free software you can then dissect one
fish out of the burrito reposition it in
some anatomically useful way do this
over and over again and for 1/2 hour CT
scan what was once an entire PhD
dissertation is now the throat clearing
for chapter 1 we can scan over 700
species per month with a single scanner
not only that the scanners here it’s
really heavy and I’m not letting anybody
take it anywhere but the data can be
processed off-site
best of all the data can be processed
with free open-source software that’s
available on every platform Linux Mac PC
whatever you want so we’re talking about
very very big data over 2,000 species
over 4000 specimens scan in the last
year because sometimes you want to know
what happens to the skeleton as
something grows so you scan 20 of the
same species we’ve scanned over 300
families of fishes and as of yesterday
we’re sitting on 50 6 terabytes of
information I can’t tell you how unhappy
I would be if those data were something
I had to encompass that I had to
surround and defend from prying eyes the
best thing about these data is they are
completely open access anyone can get
access to all of these data and analyze
it with free software anyone in the
world anywhere in the world this
completely democratizes the process of
dealing with these really critical
morphological data points so we’ve
leveraged accessibility to get lots of
people access and learn a lot more than
we ever could we put these things up on
a really neat website called Morpho
source so you can go here and download
any of the datasets and we’re already
seeing the benefits I got a very
apologetic email from a wonderful German
neurobiologist Michael Hoffman who was
who he was terribly embarrassed to tell
me that he had downloaded all of my data
and because he’s a web wizard and great
at making surfaces he had taken all of
my data and taken these slices and
turned them into 3d printable rotatable
models
and he put them up on a website for
anybody to download for any purpose and
he was very sorry because he wasn’t sure
I was getting enough credit for it
please do more terrible things like this
to me this is awesome so now you can go
to this virtual Museum of Natural
History and Michael’s put my name
embarrassingly on the front page saying
you know here’s a bunch of Adam’s fish
they’re not my fish they’re everybody’s
fish the availability of these data
allowed a computer Sciences team
interested in the concept of extracting
an object from a field of data that was
completely unknown it inspired them to
try their tools on my fish burritos and
so they produced this thing called topo
angler and they took a job that we were
doing manually over the course of six or
seven hours and dropped it down to a
15-minute task because the data were
there they could have it for any purpose
they wanted so what’s next for science
what’s next is a lot more collaboration
a lot more sharing of data more open
access data more open access publication
of scientific results so that you’re not
sitting there reading some I mean
terrible excuse for science news you can
follow a link and get to the scientific
paper and most excitingly the
availability of these huge data sets to
everybody invite synergies across fields
physicists chemists computer scientists
and biologists all realizing that we
work on very similar problems and often
need very similar tools so these 3d data
which you know for me
are useful as inspiration for new
engineering and new technologies and
biomaterials for other people might be
useful for something completely
different in fact their highest purpose
is certainly not the ideas that I have
as I’m gathering these data the highest
purposes are out there and by giving
people access we ensure that the
possibility that they fulfill these is
there so what’s next is more access more
collaboration thank you
[Applause]
hi thanks very much for coming I’m gonna
tell you a little bit about one thing
that’s next for science and I’m gonna
use an example from my own research but
what I’m talking about is broadly
applicable to science writ large I’m
gonna talk to you about very big data
that’s been set free data for my whole
career has been big I’ve always had full
diskettes now discs now raids when I
started they were 92 kilobytes and I
filled them then a hundred megabytes
then 10 gigabytes and now huge amounts
so as physiologists we looked at data in
the kilobytes to may be megabytes worth
of data genomics which has done a
fabulous job both of being big and of
sharing and being open works in the
megabytes thinking a bytes range I’m
interested in shape I’m interested in
the shape of skeletons and the soft
tissue that surrounds them and those
data are huge so they run from the
gigabytes to terabytes
for single datasets and date has always
been big relative to our capabilities
but until very very recently data was
not free data was locked into little
silos in scientists labs
where you couldn’t get at them where I
couldn’t get at them where their
colleagues couldn’t get at them where
their spouses and children couldn’t get
at them nobody could get at these data
sometimes ten years later they couldn’t
get at so I’m gonna talk to you about
big morphological data sets the data set
I’m gonna talk about is computed
tomography scans CT scans first thing I
tell you what a CT scan is which
requires some fabulous finger puppetry
behold a photorealistic bird look only
at the screen and you see the bird and
it is mysterious to you because you
didn’t see me come in with a bird but
now there’s one but if I freeze my hand
in place now look only at the screen and
rotate it through a hundred and eighty
degrees watching just those shadows in
your brain you can now not only see that
that wasn’t a bird you can put your
fingers in the same position as my
fingers were in you can deconvolve those
shadows into the 3d shape of my hand CT
scanning is exactly that except instead
of light it uses x-rays so instead of a
black and white shadow there’s a
grayscale shadow with the darkest bits
being the most radio-opaque and the
lightest bits being the least
radio-opaque
so the CT scan generates a 3d volume of
density
the end result looks like that that is a
piranha a beautiful fish that I scanned
because I wanted to see its teeth which
are the business end of the thing right
that’s what everyone wants to see
but the moment I looked at that scan I
realized I knew something about how it
lives way over on the side there you can
see two ribs that have been bisected
because this is a schooling fish that’s
really well armed and when they have
their little discussions amongst each
other they lead to serious wounds so
these models of skeletons can tell us
how fish lives and I’m interested in CT
scanning all of the fishes not every
individual but every species its 33,000
species of fish and at conventional CT
scanning rates we’re talking about a 45
year project so just about when I hit
middle age I should be done I plan to
take up exercise between now and then
the killer is this effort should cost
between 3 and 300 million dollars and as
much as I deeply love fish and I really
do I mean I’m deeply fascinated with
them I couldn’t justify spending those
kinds of funds on scanning fish
skeletons so we got a choice we can
either scan fewer species or we can scan
faster and I am fundamentally not a
fewer kind of person
so we raised some philanthropic funds
actually a very large set of donations
allowed us to buy a micro CT scanner
that lives just down the hall in my
office and from this scanner I
discovered two truths to what should
have been obvious truths but which to me
at the time
we’re not obvious at all the first of
these is that all CT scanning
interrogate s– a cylindrical volume how
many people here have actually been in a
CT scanner just give a raise van oh wow
huge numbers that’s awesome so you’ve
laid there you lay there in a bed and
either the bed goes into a doughnut or
in some models the donut runs down the
bed that produces a beautiful
three-dimensional image of the offending
part of your body what you may not know
is that everything inside the donut
everything was scanned you and your
offending part the bed the air around
all of it was scanned but we know what
air looks like and we know what beds
look like so this was terribly
inefficient we really should have piled
at least half a dozen people on there
with you run them all through it once
because everything was getting scanned
so that was the first realization the
second realization was more subtle and
it it turns out to be true that you
don’t want all of the resolution anytime
you take a picture there’s a certain
size for those pixels in CT the an
analogy is voxels that is volume
elements and all of you know when you
get a new camera what do people ask you
oh how many megapixels is that all right
that’s what they ask the first question
they don’t ask like how much was it do
you like it that you take an icepick
you know how many megapixels there are
another resolution if you own a CT
scanner that’s the first thing there so
what’s the resolution second thing they
ask is how big a thing can I put in
there as though there were some universe
where they would want to get five
thousand slices per millimeter of an
entire dead hippopotamus which would be
more data than fit on all the hard disks
in the United States right now you don’t
want all the resolution so how much do
you want I mean that’s a good question
one that we’ve said about to answer we
looked at a lot of studies and we found
that it depends totally on what you’re
doing with it and so what we had to do
was figure out what the worst case was
what what’s the most resolution you
could want and it turns out that the
most resolution you want for practical
purposes is about two hundred and fifty
slices across the head so give me a fish
I can measure the width of its head and
I can tell you what resolution you want
to scan it that means that I can pre
print cylinders that I can pack with
fish the cylinder on your right is a 30
micron cylinder 33 slices per millimeter
on your left a cylinder that’s larger
it’s a 50 micron cylinder it gives you
20 slices per millimeter and I can just
pack the fish in there this doesn’t work
with humans in healthcare it’s an
innovation I’ve proposed but it turns
out that people object dead fishes don’t
it’s one of the major reasons I work
with them
the resolution you aim for it depends on
the fish size which means when you bring
hundreds of fish to my lab to scan them
the first thing I have you do is arrange
them in size order we then pick the
specimens that are going to be scanned
at a particular resolution and we start
loading up the cylinder we can fit 10 to
20 specimens in that cylinder and when
we scan it we get this
we have reimagined that staple of the
mid week the fish burrito that is
somewhere around 20 individual species
in a group of sculpins scanned by a
graduate student named Thaddeus boo sir
Thaddeus came to the lab and and and we
developed this technique this is the
very first fish burrito that was ever
scanned I know this because it is
missing a key innovation that came about
with the second fish burrito that was
ever scanned little radio-opaque labels
so that we could tell one fish from
another because it turns out that only
fattiest could do this next trick using
free software you can then dissect one
fish out of the burrito reposition it in
some anatomically useful way do this
over and over again and for 1/2 hour CT
scan what was once an entire PhD
dissertation is now the throat clearing
for chapter 1 we can scan over 700
species per month with a single scanner
not only that the scanners here it’s
really heavy and I’m not letting anybody
take it anywhere but the data can be
processed off-site
best of all the data can be processed
with free open-source software that’s
available on every platform Linux Mac PC
whatever you want so we’re talking about
very very big data over 2,000 species
over 4000 specimens scan in the last
year because sometimes you want to know
what happens to the skeleton as
something grows so you scan 20 of the
same species we’ve scanned over 300
families of fishes and as of yesterday
we’re sitting on 50 6 terabytes of
information I can’t tell you how unhappy
I would be if those data were something
I had to encompass that I had to
surround and defend from prying eyes the
best thing about these data is they are
completely open access anyone can get
access to all of these data and analyze
it with free software anyone in the
world anywhere in the world this
completely democratizes the process of
dealing with these really critical
morphological data points so we’ve
leveraged accessibility to get lots of
people access and learn a lot more than
we ever could we put these things up on
a really neat website called Morpho
source so you can go here and download
any of the datasets and we’re already
seeing the benefits I got a very
apologetic email from a wonderful German
neurobiologist Michael Hoffman who was
who he was terribly embarrassed to tell
me that he had downloaded all of my data
and because he’s a web wizard and great
at making surfaces he had taken all of
my data and taken these slices and
turned them into 3d printable rotatable
models
and he put them up on a website for
anybody to download for any purpose and
he was very sorry because he wasn’t sure
I was getting enough credit for it
please do more terrible things like this
to me this is awesome so now you can go
to this virtual Museum of Natural
History and Michael’s put my name
embarrassingly on the front page saying
you know here’s a bunch of Adam’s fish
they’re not my fish they’re everybody’s
fish the availability of these data
allowed a computer Sciences team
interested in the concept of extracting
an object from a field of data that was
completely unknown it inspired them to
try their tools on my fish burritos and
so they produced this thing called topo
angler and they took a job that we were
doing manually over the course of six or
seven hours and dropped it down to a
15-minute task because the data were
there they could have it for any purpose
they wanted so what’s next for science
what’s next is a lot more collaboration
a lot more sharing of data more open
access data more open access publication
of scientific results so that you’re not
sitting there reading some I mean
terrible excuse for science news you can
follow a link and get to the scientific
paper and most excitingly the
availability of these huge data sets to
everybody invite synergies across fields
physicists chemists computer scientists
and biologists all realizing that we
work on very similar problems and often
need very similar tools so these 3d data
which you know for me
are useful as inspiration for new
engineering and new technologies and
biomaterials for other people might be
useful for something completely
different in fact their highest purpose
is certainly not the ideas that I have
as I’m gathering these data the highest
purposes are out there and by giving
people access we ensure that the
possibility that they fulfill these is
there so what’s next is more access more
collaboration thank you
[Applause]
Please follow and like us: