beyond health 2.0: the semantic web and intelligent systems · 2016. 3. 29. · beyond health 2.0:...
TRANSCRIPT
Beyond Health 2.0: the semantic web and intelligent
systems
Erik van Mulligen PhD
Marc Weeber PhD
Ravi Kalaputapu PhD
Erasmus University Medical Center, Rotterdam, The Netherlands
Knewco Inc, New York, United States of America
Netherlands Institute for Health Sciences (NIHES)
Netherlanss Center for BioInformatics (NBIC)
Health Users
John has already been using celebrex for a few years to
kill the pain of his reumatoïde artritis.
Recently he got stomach complaints, in particular pain.
As a preparation for his visit to the general practitioner he
is browsing the internet to find related information.
Is he able to combine information from different sources
(health sites, scientific literature, pharmaceutical sites) to
such an extent that he is well informed?
What tools would be necessary to locate and link right
information?
Rationale
To provide health information consumers (lay people, patients, scientists)
with relevant, useful (additional) information when consulting information (on
the web).
Create a semantic web on top of existing
information sources that links information
topics from different sites and databases
and with different modalities (text, video).
Assist health information consumers with
finding relevant, reliable information from
the information avalanche.
• much useful and relevant legacy data
and web pages
• semantic web technology still under
development
• approaches to overlay semantics on
the current web
• combination strategies
overlay current web with semantics layer
web 1.0 & 2.0
semantic web
mapping
Semantic Mining
Observational data
celecoxib causes upper gastrointestinal hemorrhage
celebrex causes Upper Gastro-Intestinal Bleeding
Peer reviewed data
Ontology
Diseases C001
Upper Gastro-Intestinal Bleeding
Upper gastrointestinal
hemorrhage
Drugs Celecoxib Celebrex
EHR
BioBanks
Studies
Literature
Guidelines
Protocols
Triple store
RDF/OWL
Ontology
EHR
BioBanks
Studies
Literature
Guidelines
Protocols
Nonsteroidal anti-inflammatory drugs (NSAIDs) are commonly used, but
have risks associated with their use, including significant upper
gastrointestinal tract bleeding. Older persons, persons taking
anticoagulants, and persons with a history of upper gastrointestinal tract
bleeding associated with NSAIDs are at especially high risk.
nonsteroidal anti-
inflammatory drugs
upper gastrointestinal tract
bleeding
anti-coagulants
older persons
causes increase
risk
increase
risk
increase risk
nonsteroidal anti-inflammatory drugs causes upper gastrointestinal tract bleeding Triple store
RDF/OWL older persons
anti-coagulants increase risk
upper gastrointestinal tract bleeding
upper gastrointestinal tract bleeding
EHR
BioBanks
Studies
Literature
Guidelines
Protocols
ontology development
-NCBO
-Unified Medical Language System
-SNOMED CT
OWL/RDF triple formalisms
-nano publication
-aggregation methods: association, mutual information
specific projects
-EU-ADR: detecting new side effects for drugs from observational data
-OpenPHACTS: combining triples for drug discovery
-CALBC: harmonization & alignment of different NER systems in a large corpus
-Semantic MedLine: semantic relations between entities in PubMed
Example: EU-ADR
Data extraction: periodic
Signal detection
Signal substantiation
Retrospective and prospective
signal validation
Literature
Known side
effects
Pathway
analysis
In-silico
simulation
Medical databases: 30 Million persons (IT, NL, UK, DK)
Data mining
Mapping of events
and drugs
Development of
extraction tools
Mapping web page to Semantics
• named entity recognition on the
fly, mapping term variants to
same concept
• disambiguation on the fly using
context
• identifying semantic
relations/triples relevant for
user
• identifying most relevant
entities on a page
• showing additional information
in text
celecoxib causes upper gastrointestinal hemorrhage
celebrex causes Upper Gastro-Intestinal Bleeding Triple store
RDF/OWL
Adding semantics
Health Users
John has already been using celebrex for a few years to
kill the pain of his reumatoïde artritis.
Using the semantic layer he now nows that celebrex is the
brand name for celecoxib which belongs to the family of
nons-steroidal anti-inflammatory drugs.
This family of drugs is known to cause upper gastro-
intestinal bleedings. He will ask his general practitioner
whether there are alternatives that don’t have these
particular side effects.
Requirements
A rich enough ontology and triple store that connects topics
On the fly analysis of web pages to identify health topics
Term variations
Disambiguation / page analysis
Bench marking (CALBC, I2B2, BioCreative, TREC)
Linkage with different information sources
Information available at the point of reading
Semantic Enrichment
Easy deployment
Enrichment provided by site
On demand enrichment
User monitoring / intelligent systems
Based on context highlight different topics
Populating relevant linked information, depending on context
Client-side user tracking to determine context
Business model
Advertisement
Licensing by site owner
Licensing by end-user (app store)
Open (source) architecture
Next…
Extending Context specific population of linked information
Drugs -> side effects
Disease -> treatments, guidelines
Linking with online electronic health records
Linking with social media (patient organizations, patients with same
disease, patients like me)
Deeper NLP
Thanks for your attention!
I’m happy to take questions either now, if time permits,
or per e-mail: [email protected]