Ai in Healthcare 202011 201117103639

AI and Healthcare
Paul
Agapow November 2020
Oncology R&D
The data is
complicated &
diverse
Labs, genomics,
clinical exams,
images, physical
measurements,
chemical,
health records,
other ‘omics,
observations,
medications …
7 Name 17 November 2020
What are our healthcare
problems?
Gathering information Understanding disease Developing Delivering healthcare

More and better data, What is a disease, intervention Diagnosing patients,
monitoring patients, new pathophysiological s predicting outcomes,
molecular technologies, mechanisms, biomarkers, Finding possible targets, targeted therapy, resource
imaging, devices, patient subtypes candidate molecules, allocation & optimization
integration of different running trials, analysing
modalities, EHR records trials
8 Name 17 November 2020
But what is AI / Machine Learning / Data
Science?
Other than things we talk about a lot …
Statistical modelling Machine Learning / AI
a continuum of approaches
Explicit Clear Few
No model
models assumptions assumptions
Clean & Trained from
controlled data Messy data
data
10
ML/AI is
well suited for • Complex multi-modal data
healthcare & • Often poor idea of underlying
therapy mechanism or model
development • Messy problems with messy data
• Lots of available data (caveat)
• Many healthcare questions are classical
data questions (classify, optimize,
predict)
• Healthcare should be data-driven
• Great success in other complex
11
domains Name 17 November 2020
But what are the
pitfalls?
Need more (labelled) data May require specialised Bias & interpretability
And healthcare data needs computation & skills – data never lies, but
to be handled carefully Some problems difficult to what is it telling us?
adapt to ML
12
Radiology & imaging widely used in
healthcare
• X-rays, CT, MRI, PET, sonograms …

• But interpretation is laborious
• Scope for human error
– 71% of detected lung cancers were
retrospectively found on previous scans
– 5-9% disagreement between experts
– 23% when no clinical information
supplied
• Not enough radiologists
• Not enough time
https://www.rsna.org/en/news/2019
/ May/uk-radiology-shortage
14
Ai is good at recognising things in images
• Lots of prior art

• Lots of data to train models
from
• “AI radiologist”
– would be more consistent
– faster
– could double-check or
triage
• But there’s more …
15
Radiomic analysis of medical images
Radiomics is the science of extracting quantitative
features from medical images to measure shape, intensity,
density, texture, etc. The analysis of these ‘radiomic
features’ can reveal disease characteristics that are not
readily appreciated by the naked eye.
Baseline scan Sequential scans
Specific scientific questions to address:

• Can we predict response to specific drugs from the baseline scan? i.e. duration of PFS or OS
• Can we define novel efficacy endpoints? i.e. identify quantitative changes in the image that predict overall
survival more robustly than conventional endpoints (e.g. RECIST)
• Can we get insight into toxicity? i.e. improved prediction, diagnosis or understanding of AEs such as ILD
• Can the scans provide other insights? e.g. tumour genetics, e.g. therapy resistance, e.g. POM biomarkers?
• Can we effectively combine radiomic insights with other clinical data in order to accelerate and
improve patient stratification algorithms?
AI for PD-L1 scoring in Urothelial Carcinoma
Deep learning can automatically score PD-L1 expression in Tumour cells and
Immune cells
Slide stained for PD-L1 expression Cells that were automatically detected using AI
AstraZeneca generates and has access to more data than ever
before.
Genetic & Patient-Centric Sensors & Interactive Healthcare Information Market

Genomic Data Data Smart Devices Media network Data
Post
Target Marketing
Target Discovery Pre-Clinical Clinical Commercial
Validatio Surveillance
ID n
“AI will not replace
drug hunters, but
drug hunters who
don’t use AI will be
replaced by those
who do.”
20
Name 17
-Andrew Hopkins, CEO Exscientia
AI for drug candidate selection & prioritization
https://www.bi
21 opharma-excellence.com/news/2019/6/30/artificial-intelligence-a-revolution-in-
biopharmaceutical-development
Patients are heterogenous
• Similar patient presentation can

mask vastly different molecular
machinery
• Even within a “homogenous”
condition, patients will have
different outcomes
• What are the treatment effects
for
individual patients?
•Understanding
More effective trials
these leads to:
• More effective treatment Heterogeneity in lesion change in colorectal cancer
• Insights on pathophysiology Nikodemiou et al. (2020)
22
AI enabled mining of electronic health records to
better understand diseases
COPD T2D
 Transform patients into sequences of diagnosis Topology based Patient-Patient network, identify
codes distinct subtypes of T2D
 Look for over-represented temporal pairs of codes
 Collapse pairs into trajectories of diagnoses
 Combine similar trajectories with graph similarity
Brunak et al. Nature Coms. 2016 Dudley et al. Sci. transl. Med, 2015
Data driven KOL identification and site selection
Network Analysis Federated EHRs
• Claims data is used to

map physician networks
based on patient
referrals
• Network analytics such

as PageRank algorithm
are used to determine
which physicians are
Patient referral network of most important in the
oncologists & surgeons
treating NSCLC based on network
claims data.
• Network connections are
Color represents physician
grouping. used to map existing
Size of bubble represents relationships between
physician PageRank.
oncologists & surgeons Real Time I/E analysis of Trial protocol
24
Building a external control arm from Real World Data
Patients with unmet Patients from historical
medical need trials / RWE data
Matched patients on standard of
Inclusion / care can be compared to new
Inclusion /
exclusion criteria treatment
exclusion criteria
Apply Propensity Score Matching
Single-arm trial
Access to New Medicine
Matching requires Deep data

not just Big Data
25

Ai in Healthcare 202011 201117103639

Uploaded by

Copyright:

Available Formats

Ai in Healthcare 202011 201117103639

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ai in Healthcare 202011 201117103639

Uploaded by

Copyright:

Available Formats

AI and Healthcare

Gathering information Understanding disease Developing Delivering healthcare

Statistical modelling Machine Learning / AI

• X-rays, CT, MRI, PET, sonograms …

• Lots of prior art

Specific scientific questions to address:

Genetic & Patient-Centric Sensors & Interactive Healthcare Information Market

• Similar patient presentation can

• Claims data is used to

• Network analytics such

Apply Propensity Score Matching

Access to New Medicine

Matching requires Deep data

You might also like