<s>
Artificial	B-Application
intelligence	I-Application
in	I-Application
healthcare	I-Application
is	O
an	O
overarching	O
term	O
used	O
to	O
describe	O
the	O
use	O
of	O
machine-learning	O
algorithms	O
and	O
software	O
,	O
or	O
artificial	B-Application
intelligence	I-Application
(	O
AI	B-Application
)	O
,	O
to	O
mimic	O
human	O
cognition	O
in	O
the	O
analysis	O
,	O
presentation	O
,	O
and	O
comprehension	O
of	O
complex	O
medical	O
and	O
health	O
care	O
data	O
.	O
</s>
<s>
Specifically	O
,	O
AI	B-Application
is	O
the	O
ability	O
of	O
computer	O
algorithms	O
to	O
approximate	O
conclusions	O
based	O
solely	O
on	O
input	O
data	O
.	O
</s>
<s>
The	O
primary	O
aim	O
of	O
health-related	O
AI	B-Application
applications	O
is	O
to	O
analyze	O
relationships	O
between	O
clinical	O
techniques	O
and	O
patient	O
outcomes	O
.	O
</s>
<s>
AI	B-Application
programs	O
are	O
applied	O
to	O
practices	O
such	O
as	O
diagnostics	O
,	O
treatment	O
protocol	O
development	O
,	O
drug	O
development	O
,	O
personalized	O
medicine	O
,	O
and	O
patient	O
monitoring	O
and	O
care	O
.	O
</s>
<s>
What	O
differentiates	O
AI	B-Application
technology	O
from	O
traditional	O
technologies	O
in	O
healthcare	O
is	O
the	O
ability	O
to	O
gather	O
data	O
,	O
process	O
it	O
,	O
and	O
produce	O
a	O
well-defined	O
output	O
to	O
the	O
end-user	O
.	O
</s>
<s>
AI	B-Application
does	O
this	O
through	O
machine	O
learning	O
algorithms	O
and	O
deep	B-Algorithm
learning	I-Algorithm
.	O
</s>
<s>
AI	B-Application
algorithms	O
behave	O
differently	O
from	O
humans	O
in	O
two	O
ways	O
:	O
(	O
1	O
)	O
algorithms	O
are	O
literal	O
:	O
once	O
a	O
goal	O
is	O
set	O
,	O
the	O
algorithm	O
learns	O
exclusively	O
from	O
the	O
input	O
data	O
and	O
can	O
only	O
understand	O
what	O
it	O
has	O
been	O
programmed	O
to	O
do	O
,	O
(	O
2	O
)	O
and	O
some	O
deep	B-Algorithm
learning	I-Algorithm
algorithms	O
are	O
black	B-Device
boxes	I-Device
;	O
algorithms	O
can	O
predict	O
with	O
extreme	O
precision	O
,	O
but	O
offer	O
little	O
to	O
no	O
comprehensible	O
explanation	O
to	O
the	O
logic	O
behind	O
its	O
decisions	O
aside	O
from	O
the	O
data	O
and	O
type	O
of	O
algorithm	O
used	O
.	O
</s>
<s>
As	O
widespread	O
use	O
of	O
AI	B-Application
in	I-Application
healthcare	I-Application
is	O
relatively	O
new	O
,	O
research	O
is	O
ongoing	O
into	O
its	O
application	O
in	O
various	O
fields	O
of	O
medicine	O
and	O
industry	O
.	O
</s>
<s>
Research	O
in	O
the	O
1960s	O
and	O
1970s	O
produced	O
the	O
first	O
problem-solving	O
program	O
,	O
or	O
expert	B-Application
system	I-Application
,	O
known	O
as	O
Dendral	O
.	O
</s>
<s>
While	O
it	O
was	O
designed	O
for	O
applications	O
in	O
organic	O
chemistry	O
,	O
it	O
provided	O
the	O
basis	O
for	O
a	O
subsequent	O
system	O
MYCIN	O
,	O
considered	O
one	O
of	O
the	O
most	O
significant	O
early	O
uses	O
of	O
artificial	B-Application
intelligence	I-Application
in	O
medicine	O
.	O
</s>
<s>
During	O
this	O
time	O
,	O
there	O
was	O
a	O
recognition	O
by	O
researchers	O
and	O
developers	O
that	O
AI	B-Application
systems	O
in	O
healthcare	O
must	O
be	O
designed	O
to	O
accommodate	O
the	O
absence	O
of	O
perfect	O
data	O
and	O
build	O
on	O
the	O
expertise	O
of	O
physicians	O
.	O
</s>
<s>
Approaches	O
involving	O
fuzzy	O
set	O
theory	O
,	O
Bayesian	O
networks	O
,	O
and	O
artificial	B-Architecture
neural	I-Architecture
networks	I-Architecture
,	O
have	O
been	O
applied	O
to	O
intelligent	O
computing	O
systems	O
in	O
healthcare	O
.	O
</s>
<s>
Medical	O
and	O
technological	O
advancements	O
occurring	O
over	O
this	O
half-century	O
period	O
that	O
have	O
enabled	O
the	O
growth	O
of	O
healthcare-related	O
applications	O
of	O
AI	B-Application
to	O
include	O
:	O
</s>
<s>
AI	B-Application
algorithms	O
can	O
also	O
be	O
used	O
to	O
analyze	O
large	O
amounts	O
of	O
data	O
through	O
electronic	B-Application
health	I-Application
records	I-Application
for	O
disease	O
prevention	O
and	O
diagnosis	O
.	O
</s>
<s>
Medical	O
institutions	O
such	O
as	O
The	O
Mayo	O
Clinic	O
,	O
Memorial	O
Sloan	O
Kettering	O
Cancer	O
Center	O
,	O
and	O
the	O
British	O
National	O
Health	O
Service	O
,	O
have	O
developed	O
AI	B-Application
algorithms	O
for	O
their	O
departments	O
.	O
</s>
<s>
Large	O
technology	O
companies	O
such	O
as	O
IBM	O
and	O
Google	B-Application
,	O
have	O
also	O
developed	O
AI	B-Application
algorithms	O
for	O
healthcare	O
.	O
</s>
<s>
Additionally	O
,	O
hospitals	O
are	O
looking	O
to	O
AI	B-Application
software	I-Application
to	O
support	O
operational	O
initiatives	O
that	O
increase	O
cost	O
saving	O
,	O
improve	O
patient	O
satisfaction	O
,	O
and	O
satisfy	O
their	O
staffing	O
and	O
workforce	O
needs	O
.	O
</s>
<s>
Currently	O
,	O
the	O
United	O
States	O
government	O
is	O
investing	O
billions	O
of	O
dollars	O
to	O
progress	O
the	O
development	O
of	O
AI	B-Application
in	I-Application
healthcare	I-Application
.	O
</s>
<s>
Artificial	B-Application
intelligence	I-Application
algorithms	O
have	O
shown	O
promising	O
results	O
in	O
accurately	O
diagnosing	O
and	O
risk	O
stratifying	O
patients	O
with	O
concern	O
for	O
coronary	O
artery	O
disease	O
,	O
showing	O
potential	O
as	O
an	O
initial	O
triage	O
tool	O
,	O
though	O
few	O
studies	O
have	O
directly	O
compared	O
the	O
accuracy	O
of	O
machine	O
learning	O
models	O
to	O
clinician	O
diagnostic	O
ability	O
.	O
</s>
<s>
Wearables	O
,	O
smartphones	O
,	O
and	O
internet-based	O
technologies	O
have	O
also	O
shown	O
the	O
ability	O
to	O
monitor	O
patients	O
 '	O
cardiac	O
data	O
points	O
,	O
expanding	O
the	O
amount	O
of	O
data	O
and	O
the	O
various	O
settings	O
AI	B-Application
models	O
can	O
use	O
and	O
potentially	O
enabling	O
earlier	O
detection	O
of	O
cardiac	O
events	O
occurring	O
outside	O
of	O
the	O
hospital	O
.	O
</s>
<s>
Another	O
growing	O
area	O
of	O
research	O
is	O
the	O
utility	O
of	O
AI	B-Application
in	O
classifying	O
heart	O
sounds	O
and	O
diagnosing	O
valvular	O
disease	O
.	O
</s>
<s>
Challenges	O
of	O
AI	B-Application
in	O
cardiovascular	O
medicine	O
have	O
included	O
the	O
limited	O
data	O
available	O
to	O
train	O
machine	O
learning	O
models	O
,	O
such	O
as	O
limited	O
data	O
on	O
social	O
determinants	O
of	O
health	O
as	O
they	O
pertain	O
to	O
cardiovascular	O
disease	O
.	O
</s>
<s>
Dermatology	O
is	O
an	O
imaging	O
abundant	O
speciality	O
and	O
the	O
development	O
of	O
deep	B-Algorithm
learning	I-Algorithm
has	O
been	O
strongly	O
tied	O
to	O
image	B-Algorithm
processing	I-Algorithm
.	O
</s>
<s>
Therefore	O
,	O
there	O
is	O
a	O
natural	O
fit	O
between	O
the	O
dermatology	O
and	O
deep	B-Algorithm
learning	I-Algorithm
.	O
</s>
<s>
For	O
each	O
modality	O
,	O
deep	B-Algorithm
learning	I-Algorithm
showed	O
great	O
progress	O
.	O
</s>
<s>
demonstrated	O
a	O
convolutional	B-Architecture
neural	I-Architecture
network	I-Architecture
that	O
achieved	O
94%	O
accuracy	O
at	O
identifying	O
skin	O
cells	O
from	O
microscopic	O
Tzanck	O
smear	O
images	O
.	O
</s>
<s>
Recent	O
advances	O
have	O
suggested	O
the	O
use	O
of	O
AI	B-Application
to	O
describe	O
and	O
evaluate	O
the	O
outcome	O
of	O
maxillo-facial	O
surgery	O
or	O
the	O
assessment	O
of	O
cleft	O
palate	O
therapy	O
in	O
regard	O
to	O
facial	O
attractiveness	O
or	O
age	O
appearance	O
.	O
</s>
<s>
In	O
2018	O
,	O
a	O
paper	O
published	O
in	O
the	O
journal	O
Annals	O
of	O
Oncology	O
mentioned	O
that	O
skin	O
cancer	O
could	O
be	O
detected	O
more	O
accurately	O
by	O
an	O
artificial	B-Application
intelligence	I-Application
system	O
(	O
which	O
used	O
a	O
deep	B-Algorithm
learning	I-Algorithm
convolutional	B-Architecture
neural	I-Architecture
network	I-Architecture
)	O
than	O
by	O
dermatologists	O
.	O
</s>
<s>
On	O
average	O
,	O
the	O
human	O
dermatologists	O
accurately	O
detected	O
86.6	O
%	O
of	O
skin	O
cancers	O
from	O
the	O
images	O
,	O
compared	O
to	O
95%	O
for	O
the	O
CNN	B-Architecture
machine	O
.	O
</s>
<s>
AI	B-Application
can	O
play	O
a	O
role	O
in	O
various	O
facets	O
of	O
the	O
field	O
of	O
gastroenterology	O
.	O
</s>
<s>
Endoscopic	B-Application
exams	O
such	O
as	O
esophagogastroduodenoscopies	B-Application
(	O
EGD	O
)	O
and	O
colonoscopies	O
rely	O
on	O
rapid	O
detection	O
of	O
abnormal	O
tissue	O
.	O
</s>
<s>
By	O
enhancing	O
these	O
endoscopic	B-Application
procedures	O
with	O
AI	B-Application
,	O
clinicians	O
can	O
more	O
rapidly	O
identify	O
diseases	O
,	O
determine	O
their	O
severity	O
,	O
and	O
visualize	O
blind	O
spots	O
.	O
</s>
<s>
Early	O
trials	O
in	O
using	O
AI	B-Application
detection	O
systems	O
of	O
early	O
gastric	O
cancer	O
have	O
shown	O
sensitivity	O
close	O
to	O
expert	O
endoscopists	B-Application
.	O
</s>
<s>
AI	B-Application
has	O
shown	O
potential	O
in	O
both	O
the	O
laboratory	O
and	O
clinical	O
spheres	O
of	O
infectious	O
disease	O
medicine	O
.	O
</s>
<s>
As	O
the	O
novel	O
coronavirus	O
ravages	O
through	O
the	O
globe	O
,	O
the	O
United	O
States	O
is	O
estimated	O
to	O
invest	O
more	O
than	O
$2	O
billion	O
in	O
AI-related	O
healthcare	O
research	O
by	O
2025	O
,	O
more	O
than	O
4	O
times	O
the	O
amount	O
spent	O
in	O
2019	O
( $463	O
million	O
)	O
.	O
</s>
<s>
Neural	B-Architecture
networks	I-Architecture
have	O
been	O
developed	O
to	O
rapidly	O
and	O
accurately	O
detect	O
a	O
host	O
response	O
to	O
COVID-19	O
from	O
mass	O
spectrometry	O
samples	O
.	O
</s>
<s>
Other	O
applications	O
include	O
support-vector	B-Algorithm
machines	I-Algorithm
identifying	O
antimicrobial	O
resistance	O
,	O
machine	O
learning	O
analysis	O
of	O
blood	O
smears	O
to	O
detect	O
malaria	O
,	O
and	O
improved	O
point-of-care	O
testing	O
of	O
Lyme	O
disease	O
based	O
on	O
antigen	O
detection	O
.	O
</s>
<s>
Additionally	O
,	O
AI	B-Application
has	O
been	O
investigated	O
for	O
improving	O
diagnosis	O
of	O
meningitis	O
,	O
sepsis	O
,	O
and	O
tuberculosis	O
,	O
as	O
well	O
as	O
predicting	O
treatment	O
complications	O
in	O
hepatitis	O
B	O
and	O
hepatitis	O
C	O
patients	O
.	O
</s>
<s>
AI	B-Application
has	O
been	O
explored	O
for	O
use	O
in	O
cancer	O
diagnosis	O
,	O
risk	O
stratification	O
,	O
molecular	O
characterization	O
of	O
tumors	O
,	O
and	O
cancer	O
drug	O
discovery	O
.	O
</s>
<s>
A	O
particular	O
challenge	O
in	O
oncologic	O
care	O
that	O
AI	B-Application
is	O
being	O
developed	O
to	O
address	O
is	O
the	O
ability	O
to	O
accurately	O
predict	O
which	O
treatment	O
protocols	O
will	O
be	O
best	O
suited	O
for	O
each	O
patient	O
based	O
on	O
their	O
individual	O
genetic	O
,	O
molecular	O
,	O
and	O
tumor-based	O
characteristics	O
.	O
</s>
<s>
Through	O
its	O
ability	O
to	O
translate	O
images	O
to	O
mathematical	O
sequences	O
,	O
AI	B-Application
has	O
been	O
trialed	O
in	O
cancer	O
diagnostics	O
with	O
the	O
reading	O
of	O
imaging	O
studies	O
and	O
pathology	O
slides	O
.	O
</s>
<s>
In	O
January	O
2020	O
,	O
researchers	O
demonstrated	O
an	O
AI	B-Application
system	O
,	O
based	O
on	O
a	O
Google	B-Application
DeepMind	I-Application
algorithm	O
,	O
capable	O
of	O
surpassing	O
human	O
experts	O
in	O
breast	O
cancer	O
detection	O
.	O
</s>
<s>
In	O
July	O
2020	O
,	O
it	O
was	O
reported	O
that	O
an	O
AI	B-Application
algorithm	O
developed	O
by	O
the	O
University	O
of	O
Pittsburgh	O
achieves	O
the	O
highest	O
accuracy	O
to	O
date	O
in	O
identifying	O
prostate	O
cancer	O
,	O
with	O
98%	O
sensitivity	O
and	O
97%	O
specificity	O
.	O
</s>
<s>
In	O
2018	O
,	O
the	O
U.S.	O
Food	O
and	O
Drug	O
Administration	O
authorized	O
the	O
marketing	O
of	O
the	O
first	O
medical	O
device	O
to	O
diagnose	O
a	O
specific	O
type	O
of	O
eye	O
disease	O
,	O
diabetic	O
retinopathy	O
using	O
an	O
artificial	B-Application
intelligence	I-Application
algorithm	O
.	O
</s>
<s>
Moreover	O
,	O
AI	B-Application
technology	O
may	O
be	O
utilized	O
to	O
further	O
improve	O
"	O
diagnosis	O
rates	O
"	O
because	O
of	O
the	O
potential	O
to	O
decrease	O
detection	O
time	O
.	O
</s>
<s>
AI-assisted	O
pathology	O
tools	O
have	O
been	O
developed	O
to	O
assist	O
with	O
the	O
diagnosis	O
of	O
a	O
number	O
of	O
diseases	O
,	O
including	O
breast	O
cancer	O
,	O
hepatitis	O
B	O
,	O
gastric	O
cancer	O
,	O
and	O
colorectal	O
cancer	O
.	O
</s>
<s>
AI	B-Application
has	O
also	O
been	O
used	O
to	O
predict	O
genetic	O
mutations	O
and	O
prognosticate	O
disease	O
outcomes	O
.	O
</s>
<s>
AI	B-Application
is	O
well-suited	O
for	O
use	O
in	O
low-complexity	O
pathological	O
analysis	O
of	O
large-scale	O
screening	O
samples	O
,	O
such	O
as	O
colorectal	O
or	O
breast	O
cancer	O
screening	O
,	O
thus	O
lessening	O
the	O
burden	O
on	O
pathologists	O
and	O
allowing	O
for	O
faster	O
turnaround	O
of	O
sample	O
analysis	O
.	O
</s>
<s>
Several	O
deep	B-Algorithm
learning	I-Algorithm
and	O
artificial	B-Architecture
neural	I-Architecture
network	I-Architecture
models	O
have	O
shown	O
accuracy	O
similar	O
to	O
that	O
of	O
human	O
pathologists	O
,	O
and	O
a	O
study	O
of	O
deep	B-Algorithm
learning	I-Algorithm
assistance	O
in	O
diagnosing	O
metastatic	O
breast	O
cancer	O
in	O
lymph	O
nodes	O
showed	O
that	O
the	O
accuracy	O
of	O
humans	O
with	O
the	O
assistance	O
of	O
a	O
deep	B-Algorithm
learning	I-Algorithm
program	O
was	O
higher	O
than	O
either	O
the	O
humans	O
alone	O
or	O
the	O
AI	B-Application
program	O
alone	O
.	O
</s>
<s>
Additionally	O
,	O
implementation	O
of	O
digital	O
pathology	O
is	O
predicted	O
to	O
save	O
over	O
$12	O
million	O
for	O
a	O
university	O
center	O
over	O
the	O
course	O
of	O
five	O
years	O
,	O
though	O
savings	O
attributed	O
to	O
AI	B-Application
specifically	O
have	O
not	O
yet	O
been	O
widely	O
researched	O
.	O
</s>
<s>
The	O
use	O
of	O
augmented	B-General_Concept
and	O
virtual	B-Application
reality	I-Application
could	O
prove	O
to	O
be	O
a	O
stepping	O
stone	O
to	O
wider	O
implementation	O
of	O
AI-assisted	O
pathology	O
,	O
as	O
they	O
can	O
highlight	O
areas	O
of	O
concern	O
on	O
a	O
pathology	O
sample	O
and	O
present	O
them	O
in	O
real-time	O
to	O
a	O
pathologist	O
for	O
more	O
efficient	O
review	O
.	O
</s>
<s>
AI	B-Application
also	O
has	O
the	O
potential	O
to	O
identify	O
histological	O
findings	O
at	O
levels	O
beyond	O
what	O
the	O
human	O
eye	O
can	O
see	O
,	O
and	O
has	O
shown	O
the	O
ability	O
to	O
utilize	O
genotypic	O
and	O
phenotypic	O
data	O
to	O
more	O
accurately	O
detect	O
the	O
tumor	O
of	O
origin	O
for	O
metastatic	O
cancer	O
.	O
</s>
<s>
One	O
of	O
the	O
major	O
current	O
barriers	O
to	O
widespread	O
implementation	O
of	O
AI-assisted	O
pathology	O
tools	O
is	O
the	O
lack	O
of	O
prospective	O
,	O
randomized	O
,	O
multi-center	O
controlled	O
trials	O
in	O
determining	O
the	O
true	O
clinical	O
utility	O
of	O
AI	B-Application
for	O
pathologists	O
and	O
patients	O
,	O
highlighting	O
a	O
current	O
area	O
of	O
need	O
in	O
AI	B-Application
and	I-Application
healthcare	I-Application
research	O
.	O
</s>
<s>
Primary	O
care	O
has	O
become	O
one	O
key	O
development	O
area	O
for	O
AI	B-Application
technologies	O
.	O
</s>
<s>
AI	B-Application
in	O
primary	O
care	O
has	O
been	O
used	O
for	O
supporting	O
decision	O
making	O
,	O
predictive	O
modelling	O
,	O
and	O
business	O
analytics	O
.	O
</s>
<s>
There	O
are	O
only	O
a	O
few	O
examples	O
of	O
AI	B-Application
decision	O
support	O
systems	O
that	O
were	O
prospectively	O
assessed	O
on	O
clinical	O
efficacy	O
when	O
used	O
in	O
practice	O
by	O
physicians	O
.	O
</s>
<s>
In	O
psychiatry	O
,	O
AI	B-Application
applications	O
are	O
still	O
in	O
a	O
phase	O
of	O
proof-of-concept	O
.	O
</s>
<s>
AI	B-Application
is	O
being	O
studied	O
within	O
the	O
field	O
of	O
radiology	O
to	O
detect	O
and	O
diagnose	O
diseases	O
through	O
Computerized	O
Tomography	O
(	O
CT	O
)	O
and	O
Magnetic	O
Resonance	O
(	O
MR	O
)	O
Imaging	O
.	O
</s>
<s>
Several	O
deep	B-Algorithm
learning	I-Algorithm
models	O
have	O
shown	O
the	O
capability	O
to	O
be	O
roughly	O
as	O
accurate	O
as	O
healthcare	O
professionals	O
in	O
identifying	O
diseases	O
through	O
medical	O
imaging	O
,	O
though	O
few	O
of	O
the	O
studies	O
reporting	O
these	O
findings	O
have	O
been	O
externally	O
validated	O
.	O
</s>
<s>
AI	B-Application
can	O
also	O
provide	O
non-interpretive	O
benefit	O
to	O
radiologists	O
,	O
such	O
as	O
reducing	O
noise	O
in	O
images	O
,	O
creating	O
high-quality	O
images	O
from	O
lower	O
doses	O
of	O
radiation	O
,	O
enhancing	O
MR	O
image	O
quality	O
,	O
and	O
automatically	O
assessing	O
image	O
quality	O
.	O
</s>
<s>
Further	O
research	O
investigating	O
the	O
use	O
of	O
AI	B-Application
in	O
nuclear	O
medicine	O
focuses	O
on	O
image	O
reconstruction	O
,	O
anatomical	O
landmarking	O
,	O
and	O
the	O
enablement	O
of	O
lower	O
doses	O
in	O
imaging	O
studies	O
.	O
</s>
<s>
(	O
2017	O
)	O
demonstrated	O
that	O
there	O
are	O
several	O
types	O
of	O
AI	B-Application
techniques	O
that	O
have	O
been	O
used	O
for	O
a	O
variety	O
of	O
different	O
diseases	O
,	O
such	O
as	O
support	B-Algorithm
vector	I-Algorithm
machines	I-Algorithm
,	O
neural	B-Architecture
networks	I-Architecture
,	O
and	O
decision	O
trees	O
.	O
</s>
<s>
To	O
demonstrate	O
some	O
specifics	O
for	O
disease	O
diagnosis/classification	O
there	O
are	O
two	O
different	O
techniques	O
used	O
in	O
the	O
classification	O
of	O
these	O
diseases	O
including	O
using	O
"	O
Artificial	B-Architecture
Neural	I-Architecture
Networks	I-Architecture
(	O
ANN	O
)	O
and	O
Bayesian	O
Networks	O
(	O
BN	O
)	O
"	O
.	O
</s>
<s>
Through	O
the	O
use	O
of	O
Medical	O
Learning	O
Classifiers	O
(	O
MLC	O
's	O
)	O
,	O
Artificial	B-Application
Intelligence	I-Application
has	O
been	O
able	O
to	O
substantially	O
aid	O
doctors	O
in	O
patient	O
diagnosis	O
through	O
the	O
manipulation	O
of	O
mass	O
Electronic	B-Application
Health	I-Application
Records	I-Application
(	O
EHR	O
's	O
)	O
.	O
</s>
<s>
Medical	O
conditions	O
have	O
grown	O
more	O
complex	O
,	O
and	O
with	O
a	O
vast	O
history	O
of	O
electronic	B-Application
medical	I-Application
records	I-Application
building	O
,	O
the	O
likelihood	O
of	O
case	O
duplication	O
is	O
high	O
.	O
</s>
<s>
The	O
implementation	O
of	O
AI	B-Application
to	O
not	O
only	O
help	O
find	O
similar	O
cases	O
and	O
treatments	O
,	O
but	O
also	O
factor	O
in	O
chief	O
symptoms	O
and	O
help	O
the	O
physicians	O
ask	O
the	O
most	O
appropriate	O
questions	O
helps	O
the	O
patient	O
receive	O
the	O
most	O
accurate	O
diagnosis	O
and	O
treatment	O
possible	O
.	O
</s>
<s>
The	O
increase	O
of	O
telemedicine	O
,	O
the	O
treatment	O
of	O
patients	O
remotely	O
,	O
has	O
shown	O
the	O
rise	O
of	O
possible	O
AI	B-Application
applications	O
.	O
</s>
<s>
AI	B-Application
can	O
assist	O
in	O
caring	O
for	O
patients	O
remotely	O
by	O
monitoring	O
their	O
information	O
through	O
sensors	O
.	O
</s>
<s>
The	O
information	O
can	O
be	O
compared	O
to	O
other	O
data	O
that	O
has	O
already	O
been	O
collected	O
using	O
artificial	B-Application
intelligence	I-Application
algorithms	O
that	O
alert	O
physicians	O
if	O
there	O
are	O
any	O
issues	O
to	O
be	O
aware	O
of	O
.	O
</s>
<s>
Another	O
application	O
of	O
artificial	B-Application
intelligence	I-Application
is	O
.	O
</s>
<s>
Since	O
the	O
average	O
age	O
has	O
risen	O
due	O
to	O
a	O
longer	O
life	O
expectancy	O
,	O
artificial	B-Application
intelligence	I-Application
could	O
be	O
useful	O
in	O
helping	O
take	O
care	O
of	O
older	O
populations	O
.	O
</s>
<s>
Electronic	B-Application
health	I-Application
records	I-Application
(	O
EHR	O
)	O
are	O
crucial	O
to	O
the	O
digitalization	O
and	O
information	O
spread	O
of	O
the	O
healthcare	O
industry	O
.	O
</s>
<s>
Now	O
that	O
around	O
80%	O
of	O
medical	O
practices	O
use	O
EHR	O
,	O
the	O
next	O
step	O
is	O
to	O
use	O
artificial	B-Application
intelligence	I-Application
to	O
interpret	O
the	O
records	O
and	O
provide	O
new	O
information	O
to	O
physicians	O
.	O
</s>
<s>
One	O
application	O
uses	O
natural	B-Language
language	I-Language
processing	I-Language
(	O
NLP	B-Language
)	O
to	O
make	O
more	O
succinct	O
reports	O
that	O
limit	O
the	O
variation	O
between	O
medical	O
terms	O
by	O
matching	O
similar	O
medical	O
terms	O
.	O
</s>
<s>
NLP	B-Language
algorithms	O
consolidate	O
these	O
differences	O
so	O
that	O
larger	O
datasets	O
can	O
be	O
analyzed	O
.	O
</s>
<s>
Another	O
use	O
of	O
NLP	B-Language
identifies	O
phrases	O
that	O
are	O
redundant	O
due	O
to	O
repetition	O
in	O
a	O
physician	O
's	O
notes	O
and	O
keeps	O
the	O
relevant	O
information	O
to	O
make	O
it	O
easier	O
to	O
read	O
.	O
</s>
<s>
Other	O
applications	O
use	O
concept	B-Application
processing	I-Application
to	O
analyze	O
the	O
information	O
entered	O
by	O
the	O
current	O
patient	O
's	O
doctor	O
to	O
present	O
similar	O
cases	O
and	O
help	O
the	O
physician	O
remember	O
to	O
include	O
all	O
relevant	O
details	O
.	O
</s>
<s>
Beyond	O
making	O
content	O
edits	O
to	O
an	O
EHR	O
,	O
there	O
are	O
AI	B-Application
algorithms	O
that	O
evaluate	O
an	O
individual	O
patient	O
's	O
record	O
and	O
predict	O
a	O
risk	O
for	O
a	O
disease	O
based	O
on	O
their	O
previous	O
information	O
and	O
family	O
history	O
.	O
</s>
<s>
Physicians	O
do	O
not	O
have	O
the	O
bandwidth	O
to	O
process	O
all	O
this	O
data	O
manually	O
,	O
and	O
AI	B-Application
can	O
leverage	O
this	O
data	O
to	O
assist	O
physicians	O
in	O
treating	O
their	O
patients	O
.	O
</s>
<s>
Improvements	O
in	O
natural	B-Language
language	I-Language
processing	I-Language
led	O
to	O
the	O
development	O
of	O
algorithms	O
to	O
identify	O
drug-drug	O
interactions	O
in	O
medical	O
literature	O
.	O
</s>
<s>
Other	O
algorithms	O
identify	O
drug-drug	O
interactions	O
from	O
patterns	O
in	O
user-generated	O
content	O
,	O
especially	O
electronic	B-Application
health	I-Application
records	I-Application
and/or	O
adverse	O
event	O
reports	O
.	O
</s>
<s>
Deep	B-Algorithm
learning	I-Algorithm
algorithms	O
have	O
been	O
developed	O
to	O
parse	O
these	O
reports	O
and	O
detect	O
patterns	O
that	O
imply	O
drug-drug	O
interactions	O
.	O
</s>
<s>
Greater	O
health	O
data	O
lays	O
the	O
groundwork	O
for	O
the	O
implementation	O
of	O
AI	B-Application
algorithms	O
.	O
</s>
<s>
A	O
large	O
part	O
of	O
industry	O
focus	O
of	O
implementation	O
of	O
AI	B-Application
in	O
the	O
healthcare	O
sector	O
is	O
in	O
the	O
clinical	B-Application
decision	I-Application
support	I-Application
systems	I-Application
.	O
</s>
<s>
Numerous	O
companies	O
are	O
exploring	O
the	O
possibilities	O
of	O
the	O
incorporation	O
of	O
big	B-Application
data	I-Application
in	O
the	O
healthcare	O
industry	O
.	O
</s>
<s>
The	O
following	O
are	O
examples	O
of	O
large	O
companies	O
that	O
have	O
contributed	O
to	O
AI	B-Application
algorithms	O
for	O
use	O
in	O
healthcare	O
:	O
</s>
<s>
IBM	O
's	O
Watson	B-Application
Oncology	O
is	O
in	O
development	O
at	O
Memorial	O
Sloan	O
Kettering	O
Cancer	O
Center	O
and	O
Cleveland	O
Clinic	O
.	O
</s>
<s>
IBM	O
is	O
also	O
working	O
with	O
CVS	O
Health	O
on	O
AI	B-Application
applications	O
in	O
chronic	O
disease	O
treatment	O
and	O
with	O
Johnson	O
&	O
Johnson	O
on	O
analysis	O
of	O
scientific	O
papers	O
to	O
find	O
new	O
connections	O
for	O
drug	O
development	O
.	O
</s>
<s>
In	O
May	O
2017	O
,	O
IBM	O
and	O
Rensselaer	O
Polytechnic	O
Institute	O
began	O
a	O
joint	O
project	O
entitled	O
,	O
to	O
explore	O
using	O
AI	B-Application
technology	O
to	O
enhance	O
healthcare	O
.	O
</s>
<s>
Other	O
projects	O
include	O
medical	B-Algorithm
image	I-Algorithm
analysis	I-Algorithm
of	O
tumor	O
progression	O
and	O
the	O
development	O
of	O
programmable	O
cells	O
.	O
</s>
<s>
Google	B-Application
's	I-Application
DeepMind	B-Application
platform	O
is	O
being	O
used	O
by	O
the	O
UK	O
National	O
Health	O
Service	O
to	O
detect	O
certain	O
health	O
risks	O
through	O
data	O
collected	O
via	O
a	O
mobile	O
app	O
.	O
</s>
<s>
A	O
second	O
project	O
with	O
the	O
NHS	O
involves	O
the	O
analysis	O
of	O
medical	O
images	O
collected	O
from	O
NHS	O
patients	O
to	O
develop	O
computer	B-Application
vision	I-Application
algorithms	O
to	O
detect	O
cancerous	O
tissues	O
.	O
</s>
<s>
Intel	O
's	O
venture	O
capital	O
arm	O
Intel	O
Capital	O
recently	O
invested	O
in	O
startup	O
Lumiata	O
which	O
uses	O
AI	B-Application
to	O
identify	O
at-risk	O
patients	O
and	O
develop	O
care	O
options	O
.	O
</s>
<s>
Digital	O
consultant	O
apps	O
use	O
AI	B-Application
to	O
give	O
medical	O
consultation	O
based	O
on	O
personal	O
medical	O
history	O
and	O
common	O
medical	O
knowledge	O
.	O
</s>
<s>
Entrepreneurs	O
in	O
healthcare	O
have	O
been	O
effectively	O
using	O
seven	O
business	O
model	O
archetypes	O
to	O
take	O
AI	B-Application
solution[buzzword]	O
to	O
the	O
marketplace	O
.	O
</s>
<s>
IFlytek	O
launched	O
a	O
service	O
robot	O
"	O
Xiao	O
Man	O
"	O
,	O
which	O
integrated	O
artificial	B-Application
intelligence	I-Application
technology	O
to	O
identify	O
the	O
registered	O
customer	O
and	O
provide	O
personalized	O
recommendations	O
in	O
medical	O
areas	O
.	O
</s>
<s>
The	O
Indian	O
startup	O
Haptik	B-Application
recently	O
developed	O
a	O
WhatsApp	B-Application
chatbot	O
which	O
answers	O
questions	O
associated	O
with	O
the	O
deadly	O
coronavirus	O
in	O
India	O
.	O
</s>
<s>
With	O
the	O
market	O
for	O
AI	B-Application
expanding	O
constantly	O
,	O
large	O
tech	O
companies	O
such	O
as	O
Apple	O
,	O
Google	B-Application
,	O
Amazon	O
,	O
and	O
Baidu	O
all	O
have	O
their	O
own	O
AI	B-Application
research	I-Application
divisions	O
,	O
as	O
well	O
as	O
millions	O
of	O
dollars	O
allocated	O
for	O
acquisition	O
of	O
smaller	O
AI	B-Application
based	O
companies	O
.	O
</s>
<s>
Artificial	B-Application
intelligence	I-Application
continues	O
to	O
expand	O
in	O
its	O
abilities	O
to	O
diagnose	O
more	O
people	O
accurately	O
in	O
nations	O
where	O
fewer	O
doctors	O
are	O
accessible	O
to	O
the	O
public	O
.	O
</s>
<s>
With	O
the	O
increasing	O
capabilities	O
of	O
AI	B-Application
over	O
the	O
internet	O
,	O
advanced	O
machine	O
learning	O
algorithms	O
can	O
allow	O
patients	O
to	O
get	O
accurately	O
diagnosed	O
when	O
they	O
would	O
previously	O
have	O
no	O
way	O
of	O
knowing	O
if	O
they	O
had	O
a	O
life-threatening	O
disease	O
or	O
not	O
.	O
</s>
<s>
Using	O
AI	B-Application
in	O
developing	O
nations	O
that	O
do	O
not	O
have	O
the	O
resources	O
will	O
diminish	O
the	O
need	O
for	O
outsourcing	O
and	O
can	O
improve	O
patient	O
care	O
.	O
</s>
<s>
AI	B-Application
can	O
allow	O
for	O
not	O
only	O
diagnosis	O
of	O
patient	O
in	O
areas	O
where	O
healthcare	O
is	O
scarce	O
,	O
but	O
also	O
allow	O
for	O
a	O
good	O
patient	O
experience	O
by	O
resourcing	O
files	O
to	O
find	O
the	O
best	O
treatment	O
for	O
a	O
patient	O
.	O
</s>
<s>
The	O
ability	O
of	O
AI	B-Application
to	O
adjust	O
course	O
as	O
it	O
goes	O
also	O
allows	O
the	O
patient	O
to	O
have	O
their	O
treatment	O
modified	O
based	O
on	O
what	O
works	O
for	O
them	O
;	O
a	O
level	O
of	O
individualized	O
care	O
that	O
is	O
nearly	O
non-existent	O
in	O
developing	O
countries	O
.	O
</s>
<s>
While	O
research	O
on	O
the	O
use	O
of	O
AI	B-Application
in	I-Application
healthcare	I-Application
aims	O
to	O
validate	O
its	O
efficacy	O
in	O
improving	O
patient	O
outcomes	O
before	O
its	O
broader	O
adoption	O
,	O
its	O
use	O
may	O
nonetheless	O
introduce	O
several	O
new	O
types	O
of	O
risk	O
to	O
patients	O
and	O
healthcare	O
providers	O
,	O
such	O
as	O
algorithmic	O
bias	O
,	O
Do	O
not	O
resuscitate	O
implications	O
,	O
and	O
other	O
machine	O
morality	O
issues	O
.	O
</s>
<s>
AI	B-Application
may	O
also	O
compromise	O
the	O
protection	O
of	O
patients	O
 '	O
rights	O
,	O
such	O
as	O
the	O
right	O
to	O
informed	O
consent	O
and	O
the	O
right	O
to	O
medical	O
data	O
protection	O
.	O
</s>
<s>
These	O
challenges	O
of	O
the	O
clinical	O
use	O
of	O
AI	B-Application
have	O
brought	O
about	O
a	O
potential	O
need	O
for	O
regulations	O
.	O
</s>
<s>
In	O
May	O
2016	O
,	O
the	O
White	B-Application
House	I-Application
announced	O
its	O
plan	O
to	O
host	O
a	O
series	O
of	O
workshops	O
and	O
formation	O
of	O
the	O
National	O
Science	O
and	O
Technology	O
Council	O
(	O
NSTC	O
)	O
Subcommittee	O
on	O
Machine	O
Learning	O
and	O
Artificial	B-Application
Intelligence	I-Application
.	I-Application
</s>
<s>
In	O
October	O
2016	O
,	O
the	O
group	O
published	O
The	O
National	O
Artificial	B-Application
Intelligence	I-Application
Research	I-Application
and	O
Development	O
Strategic	O
Plan	O
,	O
outlining	O
its	O
proposed	O
priorities	O
for	O
Federally-funded	O
AI	B-Application
research	I-Application
and	O
development	O
(	O
within	O
government	O
and	O
academia	O
)	O
.	O
</s>
<s>
We	O
already	O
have	O
some	O
scientists	O
who	O
know	O
artificial	B-Application
intelligence	I-Application
and	O
machine	O
learning	O
,	O
but	O
we	O
want	O
complementary	O
people	O
who	O
can	O
look	O
forward	O
and	O
see	O
how	O
this	O
technology	O
will	O
evolve.	O
"	O
</s>
<s>
The	O
joint	O
ITU-WHO	O
Focus	O
Group	O
on	O
Artificial	B-Application
Intelligence	I-Application
for	O
Health	O
(	O
FG-AI4H	O
)	O
has	O
built	O
a	O
platform	O
-	O
known	O
as	O
the	O
ITU-WHO	O
AI	B-Application
for	O
Health	O
Framework	O
-	O
for	O
the	O
testing	O
and	O
benchmarking	O
of	O
AI	B-Application
applications	O
in	O
health	O
domain	O
.	O
</s>
<s>
This	O
plan	O
lays	O
out	O
the	O
FDA	O
's	O
future	O
plans	O
for	O
regulation	O
of	O
medical	O
devices	O
that	O
would	O
include	O
artificial	B-Application
intelligence	I-Application
in	O
their	O
software	O
.	O
</s>
<s>
Tailored	O
Regulatory	O
Framework	O
for	O
Ai/M	O
:	O
-based	O
SaMD	O
,	O
2	O
.	O
</s>
<s>
According	O
to	O
the	O
U.S.	O
Department	O
of	O
Health	O
and	O
Human	O
Services	O
,	O
the	O
Office	O
for	O
Civil	O
Rights	O
(	O
OCR	O
)	O
has	O
issued	O
guidance	O
on	O
the	O
ethical	O
use	O
of	O
AI	B-Application
in	I-Application
healthcare	I-Application
.	O
</s>
<s>
Beneficence	O
requires	O
that	O
AI	B-Application
be	O
used	O
to	O
do	O
good	O
,	O
such	O
as	O
improving	O
the	O
quality	O
of	O
care	O
and	O
reducing	O
health	O
disparities	O
.	O
</s>
<s>
Non-maleficence	O
requires	O
that	O
AI	B-Application
be	O
used	O
to	O
do	O
no	O
harm	O
,	O
such	O
as	O
avoiding	O
discrimination	O
in	O
decisions	O
.	O
</s>
<s>
Finally	O
,	O
justice	O
requires	O
that	O
AI	B-Application
be	O
used	O
fairly	O
,	O
such	O
as	O
using	O
the	O
same	O
standards	O
for	O
decisions	O
no	O
matter	O
a	O
person	O
's	O
race	O
,	O
gender	O
,	O
or	O
income	O
level	O
.	O
</s>
<s>
Moreover	O
,	O
as	O
of	O
March	O
2021	O
,	O
the	O
OCR	O
hired	O
a	O
Chief	O
Artificial	B-Application
Intelligence	I-Application
Officer	O
(	O
OCAIO	O
)	O
to	O
pursue	O
the	O
"	O
implementation	O
of	O
the	O
"	O
.	O
</s>
<s>
These	O
regulations	O
require	O
healthcare	O
providers	O
to	O
follow	O
certain	O
privacy	O
rules	O
when	O
using	O
AI	B-Application
.	O
</s>
<s>
The	O
OCR	O
also	O
requires	O
healthcare	O
providers	O
to	O
keep	O
a	O
record	O
of	O
how	O
they	O
use	O
AI	B-Application
and	O
to	O
ensure	O
that	O
their	O
AI	B-Application
systems	O
are	O
secure	O
.	O
</s>
<s>
The	O
U.S.	O
is	O
not	O
the	O
only	O
country	O
to	O
develop	O
or	O
initiate	O
regulations	O
of	O
data	O
privacy	O
with	O
AI	B-Application
.	O
</s>
<s>
The	O
recommendations	O
include	O
a	O
focus	O
on	O
equality	O
and	O
non-discrimination	O
with	O
regard	O
to	O
bias	O
in	O
AI	B-Application
,	O
as	O
well	O
as	O
human	O
dignity	O
.	O
</s>
<s>
The	O
European	O
Union	O
has	O
implemented	O
the	O
General	O
Data	O
Protection	O
Regulation	O
(	O
GDPR	O
)	O
to	O
protect	O
citizens	O
 '	O
personal	O
data	O
,	O
which	O
applies	O
to	O
the	O
use	O
of	O
AI	B-Application
in	I-Application
healthcare	I-Application
.	O
</s>
<s>
In	O
addition	O
,	O
the	O
European	O
Commission	O
has	O
established	O
guidelines	O
to	O
ensure	O
the	O
ethical	O
development	O
of	O
AI	B-Application
,	O
including	O
the	O
use	O
of	O
algorithms	O
to	O
ensure	O
fairness	O
and	O
transparency	O
.	O
</s>
<s>
With	O
GDPR	O
,	O
the	O
European	O
Union	O
was	O
the	O
first	O
to	O
regulate	O
AI	B-Application
through	O
data	O
protection	O
legislation	O
.	O
</s>
<s>
The	O
Centers	O
for	O
Medicare	O
and	O
Medicaid	O
Services	O
have	O
also	O
released	O
guidelines	O
for	O
the	O
development	O
of	O
AI-based	B-Application
medical	O
applications	O
.	O
</s>
<s>
In	O
order	O
to	O
effectively	O
train	O
Machine	O
Learning	O
and	O
use	O
AI	B-Application
in	I-Application
healthcare	I-Application
,	O
massive	O
amounts	O
of	O
data	O
must	O
be	O
gathered	O
.	O
</s>
<s>
For	O
example	O
,	O
a	O
survey	O
conducted	O
in	O
the	O
UK	O
estimated	O
that	O
63%	O
of	O
the	O
population	O
is	O
uncomfortable	O
with	O
sharing	O
their	O
personal	O
data	O
in	O
order	O
to	O
improve	O
artificial	B-Application
intelligence	I-Application
technology	O
.	O
</s>
<s>
The	O
scarcity	O
of	O
real	O
,	O
accessible	O
patient	O
data	O
is	O
a	O
hindrance	O
that	O
deters	O
the	O
progress	O
of	O
developing	O
and	O
deploying	O
more	O
artificial	B-Application
intelligence	I-Application
in	I-Application
healthcare	I-Application
.	O
</s>
<s>
According	O
to	O
a	O
2019	O
study	O
,	O
AI	B-Application
can	O
replace	O
up	O
to	O
35%	O
of	O
jobs	O
in	O
the	O
UK	O
within	O
the	O
next	O
10	O
to	O
20	O
years	O
.	O
</s>
<s>
However	O
,	O
of	O
these	O
jobs	O
,	O
it	O
was	O
concluded	O
that	O
AI	B-Application
has	O
not	O
eliminated	O
any	O
healthcare	O
jobs	O
so	O
far	O
.	O
</s>
<s>
Though	O
if	O
AI	B-Application
were	O
to	O
automate	O
healthcare-related	O
jobs	O
,	O
the	O
jobs	O
most	O
susceptible	O
to	O
automation	O
would	O
be	O
those	O
dealing	O
with	O
digital	O
information	O
,	O
radiology	O
,	O
and	O
pathology	O
,	O
as	O
opposed	O
to	O
those	O
dealing	O
with	O
doctor-to-patient	O
interaction	O
.	O
</s>
<s>
It	O
is	O
expected	O
that	O
doctors	O
who	O
take	O
advantage	O
of	O
AI	B-Application
in	I-Application
healthcare	I-Application
will	O
provide	O
greater	O
quality	O
healthcare	O
than	O
doctors	O
and	O
medical	O
establishments	O
who	O
do	O
not	O
.	O
</s>
<s>
AI	B-Application
will	O
likely	O
not	O
completely	O
replace	O
healthcare	O
workers	O
but	O
rather	O
give	O
them	O
more	O
time	O
to	O
attend	O
to	O
their	O
patients	O
.	O
</s>
<s>
AI	B-Application
will	O
ultimately	O
help	O
contribute	O
to	O
the	O
progression	O
of	O
societal	O
goals	O
which	O
include	O
better	O
communication	O
,	O
improved	O
quality	O
of	O
healthcare	O
,	O
and	O
autonomy	O
.	O
</s>
<s>
Recently	O
,	O
there	O
have	O
been	O
many	O
discussions	O
between	O
healthcare	O
experts	O
in	O
terms	O
of	O
AI	B-Application
and	O
elder	O
care	O
.	O
</s>
<s>
In	O
relation	O
to	O
elder	O
care	O
,	O
AI	B-Application
bots	O
have	O
been	O
helpful	O
in	O
guiding	O
older	O
residents	O
living	O
in	O
assisted	O
living	O
with	O
entertainment	O
and	O
company	O
.	O
</s>
<s>
Since	O
AI	B-Application
makes	O
decisions	O
solely	O
on	O
the	O
data	O
it	O
receives	O
as	O
input	O
,	O
it	O
is	O
important	O
that	O
this	O
data	O
represents	O
accurate	O
patient	O
demographics	O
.	O
</s>
<s>
A	O
recent	O
scoping	O
review	O
identified	O
18	O
equity	O
challenges	O
along	O
with	O
15	O
strategies	O
that	O
can	O
be	O
implemented	O
to	O
help	O
address	O
them	O
when	O
AI	B-Application
applications	O
are	O
developed	O
using	O
many-to-many	B-Application
mapping	O
.	O
</s>
<s>
Since	O
AI	B-Application
's	O
decisions	O
are	O
a	O
direct	O
reflection	O
of	O
its	O
input	O
data	O
,	O
the	O
data	O
it	O
receives	O
must	O
have	O
accurate	O
representation	O
of	O
patient	O
demographics	O
.	O
</s>
<s>
Therefore	O
,	O
having	O
minimal	O
patient	O
data	O
on	O
minorities	O
can	O
lead	O
to	O
AI	B-Application
making	O
more	O
accurate	O
predictions	O
for	O
majority	O
populations	O
,	O
leading	O
to	O
unintended	O
worse	O
medical	O
outcomes	O
for	O
minority	O
populations	O
.	O
</s>
<s>
In	O
addition	O
to	O
biases	O
that	O
may	O
arise	O
from	O
sample	O
selection	O
,	O
different	O
clinical	O
systems	O
used	O
to	O
collect	O
data	O
may	O
also	O
impact	O
AI	B-Application
functionality	O
.	O
</s>
