How artificial intelligence revolutionizes the world of multiple myeloma
Frontiers in Hematology,
Год журнала:
2024,
Номер
3
Опубликована: Фев. 2, 2024
Multiple
myeloma
is
the
second
most
frequent
hematologic
malignancy
worldwide
with
high
morbidity
and
mortality.
Although
it
considered
an
incurable
disease,
enhanced
understanding
of
this
neoplasm
has
led
to
new
treatments,
which
have
improved
patients’
life
expectancy.
Large
amounts
data
been
generated
through
different
studies
in
settings
clinical
trials,
prospective
registries,
real-world
cohorts,
incorporated
laboratory
tests,
flow
cytometry,
molecular
markers,
cytogenetics,
diagnostic
images,
therapy
into
routine
practice.
In
review,
we
described
how
these
can
be
processed
analyzed
using
models
artificial
intelligence,
aiming
improve
accuracy
translate
benefit,
allow
a
substantial
improvement
early
diagnosis
response
evaluation,
speed
up
analyses,
reduce
labor-intensive
process
prone
operator
bias,
evaluate
greater
number
parameters
that
provide
more
precise
information.
Furthermore,
identified
intelligence
allowed
development
integrated
predict
probability
achieving
undetectable
measurable
residual
progression-free
survival,
overall
survival
leading
better
decisions,
potential
inform
on
personalized
therapy,
could
outcomes.
Overall,
revolutionize
multiple
care,
being
necessary
validate
cohorts
develop
incorporate
daily
Язык: Английский
Integrating AIPSS‐MF and molecular predictors: A comparative analysis of prognostic models for myelofibrosis
HemaSphere,
Год журнала:
2024,
Номер
8(3)
Опубликована: Март 1, 2024
Myelofibrosis
(MF)
is
a
chronic
myeloproliferative
neoplasm
that
can
manifest
as
primary
condition
(primary
myelofibrosis
[PMF])
or
after
progression
from
polycythemia
vera
essential
thrombocythemia
(secondary
[SMF]).
The
aberrant
activation
of
the
JAK-STAT
pathway
central
to
MF
pathogenesis
which
caused
by
driver
mutations
in
JAK2,
CALR,
and
MPL
genes.
These
mutations,
along
with
additional
somatic
variants
mainly
impact
epigenetic
modifiers
spliceosome
components,
shape
clinical
features
disease.1
Although
median
overall
survival
(OS)
around
6
years,
course
heterogeneous.
only
curative
strategy,
allogeneic
hematopoietic
cell
transplantation,
carries
significant
risk
early
mortality.2
It
therefore
critical
accurately
assess
transplantation
estimate
medical
therapies
determine
most
appropriate
treatment
approach
for
each
individual.3
Several
prognostic
models
are
available
categorize
patients
into
groups.4-10
Despite
their
utility,
these
have
limitations,
such
exclusive
applicability
specific
subtypes,
need
karyotypic
analysis,
may
be
challenging
due
insufficient
bone
marrow
aspiration,
reliance
on
Next
Generation
Sequencing
(NGS)
techniques
not
widely
accessible.
To
address
we
recently
conducted
study
involving
1617
60
Spanish
institutions.
In
this
study,
employed
machine
learning
(ML)
method
develop
AIPSS-MF
(Artificial
Intelligence
Prognostic
Scoring
System
Myelofibrosis;
at
https://geneticsoncohematology.com/MF/).11,
12
This
model,
relies
eight
variables
(age,
sex,
hemoglobin,
leukocytes,
platelets,
peripheral
blasts,
constitutional
symptoms,
leukoerythroblastosis),
evaluated
diagnosis,
demonstrated
robust
capability
predict
OS
leukemia-free
(LFS).
Notably,
its
predictive
accuracy
surpassed
established
like
IPSS
PMF
MYSEC-PM
SMF
patients.
One
key
advantages
ability
provide
personalized
estimates
patient.
Furthermore,
model
based
rather
than
genomic
data,
making
it
suitable
implementation
healthcare
settings.
However,
potential
improvement
our
ML
model's
incorporating
molecular
data
could
adequately
because
proportion
did
information
time.
gap,
new
including
581
GEMFIN
database
who
had
NGS
annotation.
DNA
samples
were
isolated
blood,
mostly
within
first
year
diagnosis
(58%).
Targeted
sequencing
was
performed
locally,
although
450
(77%)
cases
analyzed
9
referral
centers.
evaluating
up
56
genes,
20
consistently
across
different
panels
(missing
rate
<10%,
Supporting
Information
S1:
Table
1).
We
considered
pathogenic
likely
variant
allele
frequency
(VAF)
≥
1%.
Characteristics
outcomes
patient
cohort
shown
2.
random
forest
LFS,
focusing
genes
availability
exceeding
90%.13
First,
three
solely
results
without
taking
account
data.
initial
mere
presence
absence
gene.
Subsequently,
second
constructed
cumulative
number
per
third
focused
VAF
mutation,
aggregating
VAFs
when
multiple
affected
single
gene
comprehensive
representation.
aimed
fit
entire
optimize
prediction
precision.
metric
assessment
out-of-bag
(cross-validated)
Harrel's
c-index.
an
iterative
elimination
less
impactful
reduce
dimensionality.
ML-derived
predictors
compared
AIPSS-MF,
IPSS,
MIPSS70
scores
using
bootstrapped
c-indexes,
implementing
500
bootstrap
iterations.
Classification
myelodepletive
versus
criteria
Coltro
et
al.14
For
prediction,
considering
proved
superior
those
presence/absence
total
mutation
count
(Supporting
3).
slightly
augmented
CALR
U2AF1
Q157
mutation.
Subsequent
variable
reduction
resulted
refined
comprised
16
achieving
c-index
0.653,
named
survival.
streamlined
underscored
significance
TP53,
SRSF2,
EZH2
(Figure
1A).
parallel,
LFS
(c-index,
0.702;
3)
showed
slight
mutational
incorporated.
Unlike
declined
upon
attempting
variables,
leading
us
retain
original
greatest
EZH2,
IDH1,
U2AF1,
RUNX1,
CBL,
IDH2
1B).
Importantly,
analysis
consistent
time
lapse
between
c-indexes
0.691
0.706
predictions,
respectively.
When
comparing
performance
score
cohort,
(bootstrapped
0.812
vs.
0.649).
Combining
both
(hereafter
referred
AIPSSmol-MFSurv
model)
modest
increase
0.816
Results
remained
excluding
training
set
(177
patients),
4).
Compared
subset
annotated
(N
=
511),
yielded
highest
(0.814
0.724
0.654
MIPSS70).
Incorporating
predictor
marginally
enhanced
AIPSS-MF's
0.817)
but
notably
boosted
0.747
0.696).
findings
groups,
under
70
years
age
diagnosed
PMF,
regardless
transplant
status
5).
While
all
displayed
suboptimal
MF,
top.
addition
improve
then
integrated
creating
AIPSSmol-MFLeuk.
moderate
over
alone
(AIPSS-MF
c-index,
0.756;
AIPSSmol-MFLeuk
0.791;
Figure
Both
better
MIPSS70,
particularly
≤70
MF.
Of
note,
6).
subsequently
comparative
newly
generated
5,
10,
20-year
predictions
Grinfeld
al.'s
(blood.predict.nhs.uk).10
Predictions
cytogenetic
annotation
genetic
7).
Our
revealed
that,
isolation,
outperformed
predicting
LFS.
superiority
forecasting
greater
integrating
present
leveraged
large
comprising
academic
non-academic
institutions
Spain,
provides
realistic
reflection
real-world
system
universal
coverage.
several
methodological
limitations
require
consideration.
variety
used
constrained
potentially
overlooking
other
important
factors.
centralized
review
increases
interpretational
disparities.
Another
limitation
informative
karyotype.
Finally,
mitigated
external
dataset
cross-validating
findings,
intrinsic
internal
validation
loom.
research
has
advance
prognostics
revising
clinical-genomic
tailored
individualized
assessments.
models,
take
advantage
power
VAFs,
traditional
methods
focus
merely
count.
reinforce
role
spliceosome,
RAS
while
reducing
relevance
ASXL1
aligning
latest
field.15-17
OS.
integration
yet
improvements,
advocate
inclusion
assessments,
where
available,
refine
predictions.
holds
decision-making,
especially
determining
ideal
timing
younger
bridge
gap
practice,
developed
accessible
online
calculator
2),
(available
https://molecular-aipss-mf.prod.gemfin-env.gemfin.click/).
tool
represents
step
toward
medicine,
offering
more
accurate
management.
summary,
contributes
existing
body
knowledge
prognostication
also
paves
way
effective
strategies,
enhancing
quality
care
complex
condition.
authors
express
gratitude
doctors,
Jacob
Kathryn
Beal,
invaluable
assistance
facilitating
comparison
method.
Juan
C.
Hernández-Boluda
prepared
database.
Adrián
Mosquera-Orgueira
analysis.
Jyoti
Nangalia
calculated
according
Mosquera-Orgueira,
Manuel
Pérez-Encinas,
wrote
paper.
All
coauthors
critically
manuscript,
made
substantial
recommendations,
approved
submission
manuscript.
declare
no
conflict
interest.
Registry
initially
sponsored
grant
Novartis
Pharmaceuticals,
Inc.
scientific
board
GEMFIN.
Funding
fraction
provided
"Proyectos
de
investigación
del
SACYL",
GRS
2509/A/22.
support
request
corresponding
author.
publicly
privacy
ethical
restrictions.
supporting
restrictions
authors.
Please
note:
publisher
responsible
content
functionality
any
supplied
Any
queries
(other
missing
content)
should
directed
author
article.
Язык: Английский
AIPSS‐MF machine learning prognostic score validation in a cohort of myelofibrosis patients treated with ruxolitinib
Cancer Reports,
Год журнала:
2023,
Номер
6(10)
Опубликована: Авг. 8, 2023
In
myelofibrosis
(MF),
new
model
scores
are
continuously
proposed
to
improve
the
ability
better
identify
patients
with
worst
outcomes.
this
context,
Artificial
Intelligence
Prognostic
Scoring
System
for
Myelofibrosis
(AIPSS-MF),
and
Response
Ruxolitinib
after
6
months
(RR6)
during
ruxolitinib
(RUX)
treatment,
could
play
a
pivotal
role
in
stratifying
these
patients.We
aimed
validate
AIPSS-MF
MF
who
started
RUX
compared
standard
prognostic
at
diagnosis
RR6
of
treatment.At
diagnosis,
performs
than
widely
used
IPSS
primary
(C-index
0.636
vs.
0.596)
MYSEC-PM
secondary
0.616
0.593).
During
we
confirmed
leading
predicting
an
inadequate
response
by
JAKi
therapy
(0.682
0.571).The
score
confirms
that
it
can
adequately
stratify
subgroup
already
models,
laying
foundations
models
developed
tailored
patient
based
on
artificial
intelligence.
Язык: Английский
Optimization of diagnosis and treatment of hematological diseases via artificial intelligence
Frontiers in Medicine,
Год журнала:
2024,
Номер
11
Опубликована: Ноя. 7, 2024
Background
Optimizing
the
diagnosis
and
treatment
of
hematological
diseases
is
a
challenging
yet
crucial
research
area.
Effective
plans
typically
require
comprehensive
integration
cell
morphology,
immunology,
cytogenetics,
molecular
biology.
These
also
consider
patient-specific
factors
such
as
disease
stage,
age,
genetic
mutation
status.
With
advancement
artificial
intelligence
(AI),
more
“AI
+
medical”
application
models
are
emerging.
In
clinical
practice,
many
AI-assisted
systems
have
been
successfully
applied
to
diseases,
enhancing
precision
efficiency
offering
valuable
solutions
for
practice.
Objective
This
study
summarizes
progress
various
in
with
focus
on
their
biology
diagnosis,
well
prognosis
prediction
treatment.
Methods
Using
PubMed,
Web
Science,
other
network
search
engines,
we
conducted
literature
studies
from
past
5
years
using
main
keywords
“artificial
intelligence”
“hematological
diseases.”
We
classified
applications
AI
according
outline
summarize
current
advancements
optimizing
difficulties
challenges
promoting
standardization
this
field.
Results
can
significantly
shorten
turnaround
times,
reduce
diagnostic
costs,
accurately
predict
outcomes
through
image-recognition
technology,
genomic
data
analysis,
mining,
pattern
recognition,
personalized
medicine.
However,
several
remain,
including
lack
product
standards,
standardized
data,
medical–industrial
collaboration,
complexity
non-interpretability
systems.
addition,
regulatory
gaps
lead
privacy
issues.
Therefore,
improvements
needed
fully
leverage
potential
promote
diseases.
Conclusion
Our
results
serve
reference
point
development
offer
suggestions
further
hematology
Язык: Английский