<s>
Medical	B-Algorithm
image	I-Algorithm
computing	I-Algorithm
(	O
MIC	O
)	O
is	O
an	O
interdisciplinary	O
field	O
at	O
the	O
intersection	O
of	O
computer	B-General_Concept
science	I-General_Concept
,	O
information	O
engineering	O
,	O
electrical	O
engineering	O
,	O
physics	O
,	O
mathematics	O
and	O
medicine	O
.	O
</s>
<s>
This	O
field	O
develops	O
computational	O
and	O
mathematical	O
methods	O
for	O
solving	O
problems	O
pertaining	O
to	O
medical	B-Application
images	I-Application
and	O
their	O
use	O
for	O
biomedical	O
research	O
and	O
clinical	O
care	O
.	O
</s>
<s>
The	O
main	O
goal	O
of	O
MIC	O
is	O
to	O
extract	O
clinically	O
relevant	O
information	O
or	O
knowledge	O
from	O
medical	B-Application
images	I-Application
.	O
</s>
<s>
While	O
closely	O
related	O
to	O
the	O
field	O
of	O
medical	B-Application
imaging	I-Application
,	O
MIC	O
focuses	O
on	O
the	O
computational	O
analysis	O
of	O
the	O
images	O
,	O
not	O
their	O
acquisition	O
.	O
</s>
<s>
The	O
methods	O
can	O
be	O
grouped	O
into	O
several	O
broad	O
categories	O
:	O
image	B-Algorithm
segmentation	I-Algorithm
,	O
image	B-Algorithm
registration	I-Algorithm
,	O
image-based	O
physiological	O
modeling	O
,	O
and	O
others	O
.	O
</s>
<s>
Medical	B-Algorithm
image	I-Algorithm
computing	I-Algorithm
typically	O
operates	O
on	O
uniformly	O
sampled	O
data	O
with	O
regular	O
x-y-z	O
spatial	O
spacing	O
(	O
images	O
in	O
2D	O
and	O
volumes	O
in	O
3D	O
,	O
generically	O
referred	O
to	O
as	O
images	O
)	O
.	O
</s>
<s>
The	O
particular	O
meaning	O
of	O
the	O
data	O
at	O
the	O
sample	O
point	O
depends	O
on	O
modality	O
:	O
for	O
example	O
a	O
CT	O
acquisition	O
collects	O
radiodensity	O
values	O
,	O
while	O
an	O
MRI	O
acquisition	O
may	O
collect	O
T1	B-Algorithm
or	O
T2-weighted	O
images	O
.	O
</s>
<s>
Fan-like	O
images	O
due	O
to	O
modalities	O
such	O
as	O
curved-array	B-Architecture
ultrasound	I-Architecture
are	O
also	O
common	O
and	O
require	O
different	O
representational	O
and	O
algorithmic	O
techniques	O
to	O
process	O
.	O
</s>
<s>
Other	O
data	O
forms	O
include	O
sheared	O
images	O
due	O
to	O
during	O
acquisition	O
;	O
and	O
unstructured	B-Algorithm
meshes	I-Algorithm
,	O
such	O
as	O
hexahedral	O
and	O
tetrahedral	O
forms	O
,	O
which	O
are	O
used	O
in	O
advanced	O
biomechanical	O
analysis	O
(	O
e.g.	O
,	O
tissue	O
deformation	O
,	O
vascular	O
transport	O
,	O
bone	O
implants	O
)	O
.	O
</s>
<s>
Segmentation	B-Algorithm
is	O
the	O
process	O
of	O
partitioning	O
an	O
image	O
into	O
different	O
meaningful	O
segments	O
.	O
</s>
<s>
In	O
medical	B-Application
imaging	I-Application
,	O
these	O
segments	O
often	O
correspond	O
to	O
different	O
tissue	O
classes	O
,	O
organs	O
,	O
pathologies	O
,	O
or	O
other	O
biologically	O
relevant	O
structures	O
.	O
</s>
<s>
Medical	B-Application
image	I-Application
segmentation	B-Algorithm
is	O
made	O
difficult	O
by	O
low	O
contrast	O
,	O
noise	O
,	O
and	O
other	O
imaging	O
ambiguities	O
.	O
</s>
<s>
Although	O
there	O
are	O
many	O
computer	B-Algorithm
vision	I-Algorithm
techniques	I-Algorithm
for	I-Algorithm
image	I-Algorithm
segmentation	I-Algorithm
,	O
some	O
have	O
been	O
adapted	O
specifically	O
for	O
medical	B-Algorithm
image	I-Algorithm
computing	I-Algorithm
.	O
</s>
<s>
Atlas-Based	O
Segmentation	B-Algorithm
:	O
For	O
many	O
applications	O
,	O
a	O
clinical	O
expert	O
can	O
manually	O
label	O
several	O
images	O
;	O
segmenting	O
unseen	O
images	O
is	O
a	O
matter	O
of	O
extrapolating	O
from	O
these	O
manually	O
labeled	O
training	O
images	O
.	O
</s>
<s>
Methods	O
of	O
this	O
style	O
are	O
typically	O
referred	O
to	O
as	O
atlas-based	O
segmentation	B-Algorithm
methods	O
.	O
</s>
<s>
Atlas-based	O
methods	O
usually	O
require	O
the	O
use	O
of	O
image	B-Algorithm
registration	I-Algorithm
in	O
order	O
to	O
align	O
the	O
atlas	O
image	O
or	O
images	O
to	O
a	O
new	O
,	O
unseen	O
image	O
.	O
</s>
<s>
Shape-Based	O
Segmentation	B-Algorithm
:	O
Many	O
methods	O
parametrize	O
a	O
template	O
shape	O
for	O
a	O
given	O
structure	O
,	O
often	O
relying	O
on	O
control	O
points	O
along	O
the	O
boundary	O
.	O
</s>
<s>
Image-Based	O
segmentation	B-Algorithm
:	O
Some	O
methods	O
initiate	O
a	O
template	O
and	O
refine	O
its	O
shape	O
according	O
to	O
the	O
image	O
data	O
while	O
minimizing	O
integral	O
error	O
measures	O
,	O
like	O
the	O
Active	B-General_Concept
contour	I-General_Concept
model	I-General_Concept
and	O
its	O
variations	O
.	O
</s>
<s>
Interactive	O
Segmentation	B-Algorithm
:	O
Interactive	O
methods	O
are	O
useful	O
when	O
clinicians	O
can	O
provide	O
some	O
information	O
,	O
such	O
as	O
a	O
seed	O
region	O
or	O
rough	O
outline	O
of	O
the	O
region	O
to	O
segment	O
.	O
</s>
<s>
An	O
algorithm	O
can	O
then	O
iteratively	O
refine	O
such	O
a	O
segmentation	B-Algorithm
,	O
with	O
or	O
without	O
guidance	O
from	O
the	O
clinician	O
.	O
</s>
<s>
Manual	O
segmentation	B-Algorithm
,	O
using	O
tools	O
such	O
as	O
a	O
paint	O
brush	O
to	O
explicitly	O
define	O
the	O
tissue	O
class	O
of	O
each	O
pixel	O
,	O
remains	O
the	O
gold	O
standard	O
for	O
many	O
imaging	O
applications	O
.	O
</s>
<s>
Recently	O
,	O
principles	O
from	O
feedback	O
control	O
theory	O
have	O
been	O
incorporated	O
into	O
segmentation	B-Algorithm
,	O
which	O
give	O
the	O
user	O
much	O
greater	O
flexibility	O
and	O
allow	O
for	O
the	O
automatic	O
correction	O
of	O
errors	O
.	O
</s>
<s>
Subjective	O
surface	O
Segmentation	B-Algorithm
:	O
This	O
method	O
is	O
based	O
on	O
the	O
idea	O
of	O
evolution	O
of	O
segmentation	B-Algorithm
function	O
which	O
is	O
governed	O
by	O
an	O
advection-diffusion	O
model	O
.	O
</s>
<s>
To	O
segment	O
an	O
object	O
,	O
a	O
segmentation	B-Algorithm
seed	O
is	O
needed	O
(	O
that	O
is	O
the	O
starting	O
point	O
that	O
determines	O
the	O
approximate	O
position	O
of	O
the	O
object	O
in	O
the	O
image	O
)	O
.	O
</s>
<s>
Consequently	O
,	O
an	O
initial	O
segmentation	B-Algorithm
function	O
is	O
constructed	O
.	O
</s>
<s>
The	O
idea	O
behind	O
the	O
subjective	O
surface	O
method	O
is	O
that	O
the	O
position	O
of	O
the	O
seed	O
is	O
the	O
main	O
factor	O
determining	O
the	O
form	O
of	O
this	O
segmentation	B-Algorithm
function	O
.	O
</s>
<s>
However	O
,	O
there	O
are	O
some	O
other	O
classification	O
of	O
image	B-Algorithm
segmentation	I-Algorithm
methods	O
which	O
are	O
similar	O
to	O
above	O
categories	O
.	O
</s>
<s>
Image	B-Algorithm
registration	I-Algorithm
is	O
a	O
process	O
that	O
searches	O
for	O
the	O
correct	O
alignment	O
of	O
images	O
.	O
</s>
<s>
An	O
example	O
is	O
the	O
registration	B-Algorithm
of	O
CT	O
and	O
PET	B-Application
images	O
to	O
combine	O
structural	O
and	O
metabolic	O
information	O
(	O
see	O
figure	O
)	O
.	O
</s>
<s>
Image	B-Algorithm
registration	I-Algorithm
is	O
used	O
in	O
a	O
variety	O
of	O
medical	O
applications	O
:	O
</s>
<s>
Combining	O
complementary	O
information	O
from	O
different	O
imaging	B-Application
modalities	I-Application
.	O
</s>
<s>
In	O
contrast	O
to	O
intra-subject	O
registration	B-Algorithm
,	O
a	O
one-to-one	O
mapping	O
may	O
not	O
exist	O
between	O
subjects	O
,	O
depending	O
on	O
the	O
structural	O
variability	O
of	O
the	O
organ	O
of	O
interest	O
.	O
</s>
<s>
Inter-subject	O
registration	B-Algorithm
is	O
required	O
for	O
atlas	O
construction	O
in	O
computational	B-Algorithm
anatomy	I-Algorithm
.	O
</s>
<s>
There	O
are	O
several	O
important	O
considerations	O
when	O
performing	O
image	B-Algorithm
registration	I-Algorithm
:	O
</s>
<s>
Common	O
choices	O
are	O
rigid	B-Algorithm
,	O
affine	B-Algorithm
,	O
and	O
deformable	O
transformation	O
models	O
.	O
</s>
<s>
B-spline	B-Algorithm
and	O
thin	B-Algorithm
plate	I-Algorithm
spline	I-Algorithm
models	O
are	O
commonly	O
used	O
for	O
parameterized	O
transformation	O
fields	O
.	O
</s>
<s>
A	O
distance	O
or	O
similarity	O
function	O
is	O
used	O
to	O
quantify	O
the	O
registration	B-Algorithm
quality	O
.	O
</s>
<s>
Multi-modal	O
registration	B-Algorithm
requires	O
a	O
more	O
sophisticated	O
similarity	O
measure	O
;	O
alternatively	O
,	O
a	O
different	O
image	O
representation	O
can	O
be	O
used	O
,	O
such	O
as	O
structural	O
representations	O
or	O
registering	O
adjacent	O
anatomy	O
.	O
</s>
<s>
A	O
recent	O
study	O
employed	O
contrastive	O
coding	O
to	O
learn	O
shared	O
,	O
dense	O
image	O
representations	O
,	O
referred	O
to	O
as	O
CoMIRs	O
(	O
Contrastive	O
Multi-modal	O
Image	O
Representations	O
)	O
which	O
enabled	O
the	O
registration	B-Algorithm
of	O
multi-modal	O
images	O
where	O
existing	O
registration	B-Algorithm
methods	O
often	O
fail	O
due	O
to	O
a	O
lack	O
of	O
sufficiently	O
similar	O
image	O
structures	O
.	O
</s>
<s>
It	O
reduced	O
the	O
multi-modal	O
registration	B-Algorithm
problem	O
to	O
a	O
mono-modal	O
one	O
,	O
in	O
which	O
general	O
intensity	O
based	O
,	O
as	O
well	O
as	O
feature-based	O
,	O
registration	B-Algorithm
algorithms	O
can	O
be	O
applied	O
.	O
</s>
<s>
For	O
continuous	O
optimization	O
,	O
gradient-based	B-Algorithm
optimization	I-Algorithm
techniques	O
are	O
applied	O
to	O
improve	O
the	O
convergence	O
speed	O
.	O
</s>
<s>
Visualization	O
plays	O
several	O
key	O
roles	O
in	O
Medical	B-Algorithm
Image	I-Algorithm
Computing	I-Algorithm
.	O
</s>
<s>
Methods	O
from	O
scientific	B-Application
visualization	I-Application
are	O
used	O
to	O
understand	O
and	O
communicate	O
about	O
medical	B-Application
images	I-Application
,	O
which	O
are	O
inherently	O
spatial-temporal	O
.	O
</s>
<s>
Data	B-Application
visualization	I-Application
and	O
data	B-General_Concept
analysis	I-General_Concept
are	O
used	O
on	O
unstructured	B-Application
data	I-Application
forms	O
,	O
for	O
example	O
when	O
evaluating	O
statistical	O
measures	O
derived	O
during	O
algorithmic	O
processing	O
.	O
</s>
<s>
Direct	B-General_Concept
interaction	I-General_Concept
with	O
data	O
,	O
a	O
key	O
feature	O
of	O
the	O
visualization	O
process	O
,	O
is	O
used	O
to	O
perform	O
visual	O
queries	O
about	O
data	O
,	O
annotate	O
images	O
,	O
guide	O
segmentation	B-Algorithm
and	O
registration	B-Algorithm
processes	O
,	O
and	O
control	O
the	O
visual	O
representation	O
of	O
data	O
(	O
by	O
controlling	O
lighting	O
rendering	O
properties	O
and	O
viewing	O
parameters	O
)	O
.	O
</s>
<s>
The	O
figure	O
"	O
Visualization	O
of	O
Medical	B-Application
Imaging	I-Application
"	O
illustrates	O
several	O
types	O
of	O
visualization	O
:	O
1	O
.	O
the	O
display	O
of	O
cross-sections	O
as	O
gray	O
scale	O
images	O
;	O
2	O
.	O
reformatted	O
views	O
of	O
gray	O
scale	O
images	O
(	O
the	O
sagittal	O
view	O
in	O
this	O
example	O
has	O
a	O
different	O
orientation	O
than	O
the	O
original	O
direction	O
of	O
the	O
image	O
acquisition	O
;	O
and	O
3	O
.	O
</s>
<s>
Medical	B-Application
images	I-Application
can	O
vary	O
significantly	O
across	O
individuals	O
due	O
to	O
people	O
having	O
organs	O
of	O
different	O
shapes	O
and	O
sizes	O
.	O
</s>
<s>
Therefore	O
,	O
representing	O
medical	B-Application
images	I-Application
to	O
account	O
for	O
this	O
variability	O
is	O
crucial	O
.	O
</s>
<s>
A	O
popular	O
approach	O
to	O
represent	O
medical	B-Application
images	I-Application
is	O
through	O
the	O
use	O
of	O
one	O
or	O
more	O
atlases	O
.	O
</s>
<s>
New	O
medical	B-Application
images	I-Application
,	O
which	O
are	O
not	O
used	O
during	O
training	O
,	O
can	O
be	O
mapped	O
to	O
an	O
atlas	O
,	O
which	O
has	O
been	O
tailored	O
to	O
the	O
specific	O
application	O
,	O
such	O
as	O
segmentation	B-Algorithm
and	O
group	O
analysis	O
.	O
</s>
<s>
This	O
deformation	O
can	O
be	O
used	O
to	O
address	O
variability	O
in	O
medical	B-Application
images	I-Application
.	O
</s>
<s>
The	O
simplest	O
approach	O
is	O
to	O
model	O
medical	B-Application
images	I-Application
as	O
deformed	O
versions	O
of	O
a	O
single	O
template	O
image	O
.	O
</s>
<s>
For	O
example	O
,	O
anatomical	O
MRI	O
brain	B-Algorithm
scans	I-Algorithm
are	O
often	O
mapped	O
to	O
the	O
MNI	O
template	O
as	O
to	O
represent	O
all	O
the	O
brain	B-Algorithm
scans	I-Algorithm
in	O
common	O
coordinates	O
.	O
</s>
<s>
For	O
example	O
,	O
an	O
anatomical	O
MRI	O
brain	B-Algorithm
scan	I-Algorithm
of	O
a	O
patient	O
with	O
severe	O
brain	O
abnormalities	O
(	O
i.e.	O
,	O
a	O
tumor	O
or	O
surgical	O
procedure	O
)	O
,	O
may	O
not	O
easily	O
map	O
to	O
the	O
MNI	O
template	O
.	O
</s>
<s>
Statistical	O
methods	O
combine	O
the	O
medical	B-Application
imaging	I-Application
field	O
with	O
modern	O
Computer	B-Application
Vision	I-Application
,	O
Machine	O
Learning	O
and	O
Pattern	O
Recognition	O
.	O
</s>
<s>
Variation	O
caused	O
by	O
the	O
disease	O
can	O
manifest	O
itself	O
as	O
abnormal	O
deformation	O
of	O
anatomy	O
(	O
see	O
Voxel-based	B-Algorithm
morphometry	I-Algorithm
)	O
.	O
</s>
<s>
Additionally	O
,	O
changes	O
in	O
biochemical	O
(	O
functional	O
)	O
activity	O
can	O
be	O
observed	O
using	O
imaging	B-Application
modalities	I-Application
such	O
as	O
Positron	B-Application
Emission	I-Application
Tomography	I-Application
.	O
</s>
<s>
The	O
comparison	O
between	O
groups	O
is	O
usually	O
conducted	O
on	O
the	O
voxel	B-Algorithm
level	O
.	O
</s>
<s>
Hence	O
,	O
the	O
most	O
popular	O
pre-processing	O
pipeline	O
,	O
particularly	O
in	O
neuroimaging	B-Algorithm
,	O
transforms	O
all	O
of	O
the	O
images	O
in	O
a	O
dataset	O
to	O
a	O
common	O
coordinate	O
frame	O
via	O
(	O
Medical	B-Application
Image	I-Application
Registration	B-Algorithm
)	O
in	O
order	O
to	O
maintain	O
correspondence	O
between	O
voxels	B-Algorithm
.	O
</s>
<s>
Given	O
this	O
voxel-wise	O
correspondence	O
,	O
the	O
most	O
common	O
Frequentist	B-General_Concept
method	O
is	O
to	O
extract	O
a	O
statistic	O
for	O
each	O
voxel	B-Algorithm
(	O
for	O
example	O
,	O
the	O
mean	O
voxel	B-Algorithm
intensity	O
for	O
each	O
group	O
)	O
and	O
perform	O
statistical	O
hypothesis	O
testing	O
to	O
evaluate	O
whether	O
a	O
null	O
hypothesis	O
is	O
or	O
is	O
not	O
supported	O
.	O
</s>
<s>
The	O
null	O
hypothesis	O
typically	O
assumes	O
that	O
the	O
two	O
cohorts	O
are	O
drawn	O
from	O
the	O
same	O
distribution	O
,	O
and	O
hence	O
,	O
should	O
have	O
the	O
same	O
statistical	O
properties	O
(	O
for	O
example	O
,	O
the	O
mean	O
values	O
of	O
two	O
groups	O
are	O
equal	O
for	O
a	O
particular	O
voxel	B-Algorithm
)	O
.	O
</s>
<s>
Since	O
medical	B-Application
images	I-Application
contain	O
large	O
numbers	O
of	O
voxels	B-Algorithm
,	O
the	O
issue	O
of	O
multiple	B-General_Concept
comparison	I-General_Concept
needs	O
to	O
be	O
addressed	O
,	O
.	O
</s>
<s>
From	O
methodological	O
point	O
of	O
view	O
,	O
current	O
techniques	O
varies	O
from	O
applying	O
standard	O
machine	O
learning	O
algorithms	O
to	O
medical	B-Application
imaging	I-Application
datasets	O
(	O
e.g.	O
</s>
<s>
Support	B-Algorithm
Vector	I-Algorithm
Machine	I-Algorithm
)	O
,	O
to	O
developing	O
new	O
approaches	O
adapted	O
for	O
the	O
needs	O
of	O
the	O
field	O
.	O
</s>
<s>
Small	O
sample	O
size	O
(	O
Curse	B-Algorithm
of	I-Algorithm
Dimensionality	I-Algorithm
)	O
:	O
a	O
large	O
medical	B-Application
imaging	I-Application
dataset	O
contains	O
hundreds	O
to	O
thousands	O
of	O
images	O
,	O
whereas	O
the	O
number	O
of	O
voxels	B-Algorithm
in	O
a	O
typical	O
volumetric	O
image	O
can	O
easily	O
go	O
beyond	O
millions	O
.	O
</s>
<s>
A	O
remedy	O
to	O
this	O
problem	O
is	O
to	O
reduce	O
the	O
number	O
of	O
features	O
in	O
an	O
informative	O
sense	O
(	O
see	O
dimensionality	B-Algorithm
reduction	I-Algorithm
)	O
.	O
</s>
<s>
Alternative	O
methods	O
based	O
on	O
feature	B-General_Concept
selection	I-General_Concept
have	O
been	O
proposed	O
,	O
.	O
</s>
<s>
Additionally	O
,	O
some	O
diseases	O
(	O
e.g.	O
,	O
autism	O
spectrum	O
disorder	O
(	O
ASD	O
)	O
,	O
schizophrenia	B-Application
,	O
mild	O
cognitive	O
impairment	O
(	O
MCI	O
)	O
)	O
can	O
be	O
characterized	O
by	O
a	O
continuous	O
or	O
nearly-continuous	O
spectra	O
from	O
mild	O
cognitive	O
impairment	O
to	O
very	O
pronounced	O
pathological	O
changes	O
.	O
</s>
<s>
Shape	O
Analysis	O
is	O
the	O
field	O
of	O
Medical	B-Algorithm
Image	I-Algorithm
Computing	I-Algorithm
that	O
studies	O
geometrical	O
properties	O
of	O
structures	O
obtained	O
from	O
different	O
imaging	B-Application
modalities	I-Application
.	O
</s>
<s>
This	O
information	O
can	O
be	O
incorporated	O
both	O
into	O
the	O
image	B-General_Concept
analysis	I-General_Concept
,	O
as	O
well	O
as	O
into	O
the	O
statistical	O
modeling	O
.	O
</s>
<s>
In	O
longitudinal	O
image	O
processing	O
,	O
segmentation	B-Algorithm
and	O
analysis	O
methods	O
of	O
individual	O
time	O
points	O
are	O
informed	O
and	O
regularized	O
with	O
common	O
information	O
usually	O
from	O
a	O
within-subject	O
template	O
.	O
</s>
<s>
Traditionally	O
,	O
medical	B-Algorithm
image	I-Algorithm
computing	I-Algorithm
has	O
seen	O
to	O
address	O
the	O
quantification	O
and	O
fusion	O
of	O
structural	O
or	O
functional	O
information	O
available	O
at	O
the	O
point	O
and	O
time	O
of	O
image	O
acquisition	O
.	O
</s>
<s>
In	O
this	O
context	O
,	O
medical	B-Application
imaging	I-Application
and	O
image	O
computing	O
play	O
an	O
increasingly	O
important	O
role	O
as	O
they	O
provide	O
systems	O
and	O
methods	O
to	O
image	O
,	O
quantify	O
and	O
fuse	O
both	O
structural	O
and	O
functional	O
information	O
about	O
the	O
human	O
being	O
in	O
vivo	O
.	O
</s>
<s>
These	O
include	O
approaches	O
based	O
on	O
partial	O
differential	O
equations	O
(	O
PDEs	O
)	O
and	O
curvature	O
driven	O
flows	O
for	O
enhancement	O
,	O
segmentation	B-Algorithm
,	O
and	O
registration	B-Algorithm
.	O
</s>
<s>
Accordingly	O
,	O
very	O
recently	O
ideas	O
from	O
control	O
have	O
recently	O
made	O
their	O
way	O
into	O
interactive	O
methods	O
,	O
especially	O
segmentation	B-Algorithm
.	O
</s>
<s>
Moreover	O
,	O
because	O
of	O
noise	O
and	O
the	O
need	O
for	O
statistical	O
estimation	O
techniques	O
for	O
more	O
dynamically	O
changing	O
imagery	O
,	O
the	O
Kalman	O
filter	O
and	O
particle	B-Algorithm
filter	I-Algorithm
have	O
come	O
into	O
use	O
.	O
</s>
<s>
Some	O
imaging	B-Application
modalities	I-Application
provide	O
very	O
specialized	O
information	O
.	O
</s>
<s>
The	O
resulting	O
images	O
cannot	O
be	O
treated	O
as	O
regular	O
scalar	O
images	O
and	O
give	O
rise	O
to	O
new	O
sub-areas	O
of	O
Medical	B-Algorithm
Image	I-Algorithm
Computing	I-Algorithm
.	O
</s>
<s>
Examples	O
include	O
diffusion	B-Algorithm
MRI	I-Algorithm
,	O
</s>
<s>
functional	B-Algorithm
MRI	I-Algorithm
and	O
others	O
.	O
</s>
<s>
Diffusion	B-Algorithm
MRI	I-Algorithm
is	O
a	O
structural	O
magnetic	O
resonance	O
imaging	O
modality	O
that	O
allows	O
measurement	O
of	O
the	O
diffusion	O
process	O
of	O
molecules	O
.	O
</s>
<s>
As	O
each	O
acquisition	O
is	O
associated	O
with	O
multiple	O
volumes	O
,	O
diffusion	B-Algorithm
MRI	I-Algorithm
has	O
created	O
a	O
variety	O
of	O
unique	O
challenges	O
in	O
medical	B-Algorithm
image	I-Algorithm
computing	I-Algorithm
.	O
</s>
<s>
In	O
medicine	O
,	O
there	O
are	O
two	O
major	O
computational	O
goals	O
in	O
diffusion	B-Algorithm
MRI	I-Algorithm
:	O
</s>
<s>
The	O
diffusion	B-Algorithm
tensor	I-Algorithm
,	O
a	O
3	O
×	O
3	O
symmetric	O
positive-definite	B-Algorithm
matrix	I-Algorithm
,	O
offers	O
a	O
straightforward	O
solution	O
to	O
both	O
of	O
these	O
goals	O
.	O
</s>
<s>
Due	O
to	O
the	O
simplicity	O
of	O
this	O
model	O
,	O
a	O
maximum	O
likelihood	O
estimate	O
of	O
the	O
diffusion	B-Algorithm
tensor	I-Algorithm
can	O
be	O
found	O
by	O
simply	O
solving	O
a	O
system	O
of	O
linear	O
equations	O
at	O
each	O
location	O
independently	O
.	O
</s>
<s>
However	O
,	O
as	O
the	O
volume	O
is	O
assumed	O
to	O
contain	O
contiguous	O
tissue	O
fibers	O
,	O
it	O
may	O
be	O
preferable	O
to	O
estimate	O
the	O
volume	O
of	O
diffusion	B-Algorithm
tensors	I-Algorithm
in	O
its	O
entirety	O
by	O
imposing	O
regularity	O
conditions	O
on	O
the	O
underlying	O
field	O
of	O
tensors	O
.	O
</s>
<s>
Scalar	O
values	O
can	O
be	O
extracted	O
from	O
the	O
diffusion	B-Algorithm
tensor	I-Algorithm
,	O
such	O
as	O
the	O
fractional	B-Algorithm
anisotropy	I-Algorithm
,	O
mean	O
,	O
axial	O
and	O
radial	O
diffusivities	O
,	O
which	O
indirectly	O
measure	O
tissue	O
properties	O
such	O
as	O
the	O
dysmyelination	O
of	O
axonal	O
fibers	O
or	O
the	O
presence	O
of	O
edema	O
.	O
</s>
<s>
Standard	O
scalar	O
image	O
computing	O
methods	O
,	O
such	O
as	O
registration	B-Algorithm
and	O
segmentation	B-Algorithm
,	O
can	O
be	O
applied	O
directly	O
to	O
volumes	O
of	O
such	O
scalar	O
values	O
.	O
</s>
<s>
However	O
,	O
to	O
fully	O
exploit	O
the	O
information	O
in	O
the	O
diffusion	B-Algorithm
tensor	I-Algorithm
,	O
these	O
methods	O
have	O
been	O
adapted	O
to	O
account	O
for	O
tensor	O
valued	O
volumes	O
when	O
performing	O
registration	B-Algorithm
and	O
segmentation	B-Algorithm
.	O
</s>
<s>
Given	O
the	O
principal	O
direction	O
of	O
diffusion	O
at	O
each	O
location	O
in	O
the	O
volume	O
,	O
it	O
is	O
possible	O
to	O
estimate	O
the	O
global	O
pathways	O
of	O
diffusion	O
through	O
a	O
process	O
known	O
as	O
tractography	B-Algorithm
.	O
</s>
<s>
However	O
,	O
due	O
to	O
the	O
relatively	O
low	O
resolution	O
of	O
diffusion	B-Algorithm
MRI	I-Algorithm
,	O
many	O
of	O
these	O
pathways	O
may	O
cross	O
,	O
kiss	O
or	O
fan	O
at	O
a	O
single	O
location	O
.	O
</s>
<s>
In	O
this	O
situation	O
,	O
the	O
single	O
principal	O
direction	O
of	O
the	O
diffusion	B-Algorithm
tensor	I-Algorithm
is	O
not	O
an	O
appropriate	O
model	O
for	O
the	O
local	O
diffusion	O
distribution	O
.	O
</s>
<s>
These	O
include	O
mixtures	O
of	O
diffusion	B-Algorithm
tensors	I-Algorithm
,	O
Q-ball	O
imaging	O
,	O
diffusion	O
spectrum	O
imaging	O
and	O
fiber	O
orientation	O
distribution	O
functions	O
,	O
which	O
typically	O
require	O
HARDI	O
acquisition	O
with	O
a	O
large	O
number	O
of	O
gradient	O
directions	O
.	O
</s>
<s>
As	O
with	O
the	O
diffusion	B-Algorithm
tensor	I-Algorithm
,	O
volumes	O
valued	O
with	O
these	O
complex	O
models	O
require	O
special	O
treatment	O
when	O
applying	O
image	O
computing	O
methods	O
,	O
such	O
as	O
registration	B-Algorithm
and	O
segmentation	B-Algorithm
.	O
</s>
<s>
Functional	B-Algorithm
magnetic	I-Algorithm
resonance	I-Algorithm
imaging	I-Algorithm
(	O
fMRI	B-Algorithm
)	O
is	O
a	O
medical	B-Application
imaging	I-Application
modality	O
that	O
indirectly	O
measures	O
neural	O
activity	O
by	O
observing	O
the	O
local	O
hemodynamics	O
,	O
or	O
blood	O
oxygen	O
level	O
dependent	O
signal	O
(	O
BOLD	O
)	O
.	O
</s>
<s>
fMRI	B-Algorithm
data	O
offers	O
a	O
range	O
of	O
insights	O
,	O
and	O
can	O
be	O
roughly	O
divided	O
into	O
two	O
categories	O
:	O
</s>
<s>
Task	O
related	O
fMRI	B-Algorithm
is	O
acquired	O
as	O
the	O
subject	O
is	O
performing	O
a	O
sequence	O
of	O
timed	O
experimental	O
conditions	O
.	O
</s>
<s>
Resting	B-Algorithm
state	I-Algorithm
fMRI	I-Algorithm
is	O
acquired	O
in	O
the	O
absence	O
of	O
any	O
experimental	O
task	O
.	O
</s>
<s>
Most	O
studies	O
of	O
resting	B-Algorithm
state	I-Algorithm
fMRI	I-Algorithm
focus	O
on	O
low	O
frequency	O
fluctuations	O
of	O
the	O
fMRI	B-Algorithm
signal	O
(	O
LF-BOLD	O
)	O
.	O
</s>
<s>
There	O
is	O
a	O
rich	O
set	O
of	O
methodology	O
used	O
to	O
analyze	O
functional	O
neuroimaging	B-Algorithm
data	O
,	O
and	O
there	O
is	O
often	O
no	O
consensus	O
regarding	O
the	O
best	O
method	O
.	O
</s>
<s>
Massive	O
univariate	O
approaches	O
that	O
probe	O
individual	O
voxels	B-Algorithm
in	O
the	O
imaging	O
data	O
for	O
a	O
relationship	O
to	O
the	O
experiment	O
condition	O
.	O
</s>
<s>
Multivariate	O
-	O
and	O
classifier	O
based	O
approaches	O
,	O
often	O
referred	O
to	O
as	O
multi	O
voxel	B-Algorithm
pattern	O
analysis	O
or	O
multi-variate	O
pattern	O
analysis	O
probe	O
the	O
data	O
for	O
global	O
and	O
potentially	O
distributed	O
responses	O
to	O
an	O
experimental	O
condition	O
.	O
</s>
<s>
Early	O
approaches	O
used	O
support	B-Algorithm
vector	I-Algorithm
machines	I-Algorithm
(	O
SVM	B-Algorithm
)	O
to	O
study	O
responses	O
to	O
visual	O
stimuli	O
.	O
</s>
<s>
The	O
majority	O
of	O
such	O
studies	O
focus	O
on	O
resting	B-Algorithm
state	I-Algorithm
data	O
to	O
parcelate	O
the	O
brain	O
or	O
to	O
find	O
correlates	O
to	O
behavioral	O
measures	O
.	O
</s>
<s>
When	O
working	O
with	O
large	O
cohorts	O
of	O
subjects	O
,	O
the	O
normalization	O
(	O
registration	B-Algorithm
)	O
of	O
individual	O
subjects	O
into	O
a	O
common	O
reference	O
frame	O
is	O
crucial	O
.	O
</s>
<s>
A	O
body	O
of	O
work	O
and	O
tools	O
exist	O
to	O
perform	O
normalization	O
based	O
on	O
anatomy	O
(	O
FSL	B-Algorithm
,	O
FreeSurfer	B-Algorithm
,	O
SPM	O
)	O
.	O
</s>
<s>
Examples	O
are	O
the	O
alignment	O
of	O
the	O
cortex	O
based	O
on	O
fMRI	B-Algorithm
signal	O
correlation	O
,	O
the	O
alignment	O
based	O
on	O
the	O
global	O
functional	O
connectivity	O
structure	O
both	O
in	O
task-	O
,	O
or	O
resting	B-Algorithm
state	I-Algorithm
data	O
,	O
and	O
the	O
alignment	O
based	O
on	O
stimulus	O
specific	O
activation	O
profiles	O
of	O
individual	O
voxels	B-Algorithm
.	O
</s>
<s>
Software	O
for	O
medical	B-Algorithm
image	I-Algorithm
computing	I-Algorithm
is	O
a	O
complex	O
combination	O
of	O
systems	O
providing	O
IO	O
,	O
visualization	O
and	O
interaction	O
,	O
user	O
interface	O
,	O
data	O
management	O
and	O
computation	O
.	O
</s>
<s>
Medical	B-Algorithm
Image	I-Algorithm
Computing	I-Algorithm
is	O
also	O
related	O
to	O
the	O
field	O
of	O
Computer	B-Application
Vision	I-Application
.	O
</s>
<s>
Proceedings	O
for	O
this	O
conference	O
are	O
published	O
by	O
Springer	O
in	O
the	O
Lecture	O
Notes	O
in	O
Computer	B-General_Concept
Science	I-General_Concept
series	O
.	O
</s>
