Lecture 1: Introduction
● Artificial Intelligence: intelligence demonstrated by machines, in contrast to the natural
intelligence displayed by humans and animals.
○ Aim: mimic human cognitive functions
● AI is hot and happening:
○ Huge amounts of data are generated with the healthcare sector
○ Increased availability of fast processors and computers
○ Improved algorithms
○ Programming has become easier
● Main types of AI within healthcare
○ Machine Learning: classify patients traits or predict probability of a specific
disease outcome
■ Deep learning: mimics operation of the human brain using multiple layers
of artificial neural networks to generate automated predictions from
training datasets
● Useful in image recognition
○ Natural language processing: these methods
extract information from unstructured data to
supplement enriched structured medical data. Turn
text into machine readable data which then can be
analysed by machine learning techniques.
● From clinical data generation to natural language
processing data enrichment, to machine learning data
analysis to clinical decision making
● AI applications in healthcare
○ Examples:
■ Help consumers to maintain their health:
fitbit
■ Decision making
■ Robots for elderly care
● Examples of AI in healthcare
○ Prediction
○ Support of diagnostic processes
○ Treatment
○ Scheduling of operating room capacity
○ Improving clinical trials of new drugs
○ Enabling precision medicine
● Challenges for the use of AI in healthcare
○ Availability of data
■ Sufficient large datasets (for complex disorders even larger dataset)
● Linking of data necessary
■ Diverse data needed
1
, ○ Algorithms may be flawed
■ Validation before implementation necessary
■ Regulatory requirements for approval → AI software considered as a
medical device
○ Privacy and security of data
○ Implementation of AI technologies
■ Patient willingness
■ Adoption by healthcare providers
● Diverse data
○ If not, predictions are not valid because specific groups are not included
○ Minorities (socioeconomic status, ethnicity) often not sufficiently represent the
data
○ Access to healthcare may be limited for minorities
● Future of AI in healthcare
○ Fundamental aspects of the Dutch healthcare system
■ Quality
■ Accessibility
■ Solidarity
● Big question: how can we use AI without compromising health(care)?
● Overview of topics in AI in medicine
○ Review paper over the past 30 years of international conference on Artificial
Intelligence in Medicine (AIME)
○ Taxonomy of research themes and topics
○ Observations:
■ Major shift in knowledge-based method to data-driven method (ML)
■ Other topics are more stable (e.g., uncertainty management, image and
signal processing, natural language processing)
■ Guideline/protocol papers most highly cited
● AIME conference
○ Started in 1985, every two year
○ Connected to “Artificial Intelligence in MEDICINE” journal (AIIM journal elsevier)
○ Focus: methods and techniques from computer science and artificial intelligence
for application in biomedicine and healthcare
● Future directions
○ Big data and personalized medicine
■ Uptake electronic health record (EHR)
■ Advances in genome sequencing technology and imaging
■ Health related data through mobile and wearable devices
○ Evidence based medicine
■ Guideline development, dissemination and implementation are still paper
based
■ New evidence grows faster and faster
2
, ■ EHR data can help for streamlining the pathway from evidence production
to clinical decision support
○ Business process meddling and process mining
○ NLP, social media and the web
● Major changes last decades
○ Medical side
■ Evidence based medicine and clinical guidelines have become leading
paradigm for clinical
decision making
■ Growing incidence of
chronic illness
■ Rising cost of healthcare
○ Technology developments
■ Increase in capacity
(computation & storage)
■ Connected world
■ New AI techniques
Lecture 2: Ontologies Motivation
● What is an ontology?
○ Addresses questions such as
■ What does it mean to be?
■ What constitutes the identity of an object?
■ What categories can we sort existing things into?
○ Ontologies , when communicated to others, foster a shared understanding of
things
● Early adopters of biomedical ontologies
○ Aristotle (382-322 BC)
■ First systematic taxonomy of biology
■ Classification of organisms by shared properties
■ Use binomial genus-differentia nomenclature
○ Galen (130-210 AD)
■ Systematic description of diseases, signs and symptoms
■ In De Febrium Differentia description of fever symptoms he uses the
Aristotelian genus-differentia approach
● Genus-differentia nomenclature
○ Genus-differentia are definitions are key to good ontologies
○ A type of definition where necessary and sufficient conditions are specified that
are composed of two parts
■ Genus: serves as a basis for a new definition; all definitions with the same
genus are considered members of that genus
■ Differentia: the portion of the definition that is not provided by the genus
3
● Artificial Intelligence: intelligence demonstrated by machines, in contrast to the natural
intelligence displayed by humans and animals.
○ Aim: mimic human cognitive functions
● AI is hot and happening:
○ Huge amounts of data are generated with the healthcare sector
○ Increased availability of fast processors and computers
○ Improved algorithms
○ Programming has become easier
● Main types of AI within healthcare
○ Machine Learning: classify patients traits or predict probability of a specific
disease outcome
■ Deep learning: mimics operation of the human brain using multiple layers
of artificial neural networks to generate automated predictions from
training datasets
● Useful in image recognition
○ Natural language processing: these methods
extract information from unstructured data to
supplement enriched structured medical data. Turn
text into machine readable data which then can be
analysed by machine learning techniques.
● From clinical data generation to natural language
processing data enrichment, to machine learning data
analysis to clinical decision making
● AI applications in healthcare
○ Examples:
■ Help consumers to maintain their health:
fitbit
■ Decision making
■ Robots for elderly care
● Examples of AI in healthcare
○ Prediction
○ Support of diagnostic processes
○ Treatment
○ Scheduling of operating room capacity
○ Improving clinical trials of new drugs
○ Enabling precision medicine
● Challenges for the use of AI in healthcare
○ Availability of data
■ Sufficient large datasets (for complex disorders even larger dataset)
● Linking of data necessary
■ Diverse data needed
1
, ○ Algorithms may be flawed
■ Validation before implementation necessary
■ Regulatory requirements for approval → AI software considered as a
medical device
○ Privacy and security of data
○ Implementation of AI technologies
■ Patient willingness
■ Adoption by healthcare providers
● Diverse data
○ If not, predictions are not valid because specific groups are not included
○ Minorities (socioeconomic status, ethnicity) often not sufficiently represent the
data
○ Access to healthcare may be limited for minorities
● Future of AI in healthcare
○ Fundamental aspects of the Dutch healthcare system
■ Quality
■ Accessibility
■ Solidarity
● Big question: how can we use AI without compromising health(care)?
● Overview of topics in AI in medicine
○ Review paper over the past 30 years of international conference on Artificial
Intelligence in Medicine (AIME)
○ Taxonomy of research themes and topics
○ Observations:
■ Major shift in knowledge-based method to data-driven method (ML)
■ Other topics are more stable (e.g., uncertainty management, image and
signal processing, natural language processing)
■ Guideline/protocol papers most highly cited
● AIME conference
○ Started in 1985, every two year
○ Connected to “Artificial Intelligence in MEDICINE” journal (AIIM journal elsevier)
○ Focus: methods and techniques from computer science and artificial intelligence
for application in biomedicine and healthcare
● Future directions
○ Big data and personalized medicine
■ Uptake electronic health record (EHR)
■ Advances in genome sequencing technology and imaging
■ Health related data through mobile and wearable devices
○ Evidence based medicine
■ Guideline development, dissemination and implementation are still paper
based
■ New evidence grows faster and faster
2
, ■ EHR data can help for streamlining the pathway from evidence production
to clinical decision support
○ Business process meddling and process mining
○ NLP, social media and the web
● Major changes last decades
○ Medical side
■ Evidence based medicine and clinical guidelines have become leading
paradigm for clinical
decision making
■ Growing incidence of
chronic illness
■ Rising cost of healthcare
○ Technology developments
■ Increase in capacity
(computation & storage)
■ Connected world
■ New AI techniques
Lecture 2: Ontologies Motivation
● What is an ontology?
○ Addresses questions such as
■ What does it mean to be?
■ What constitutes the identity of an object?
■ What categories can we sort existing things into?
○ Ontologies , when communicated to others, foster a shared understanding of
things
● Early adopters of biomedical ontologies
○ Aristotle (382-322 BC)
■ First systematic taxonomy of biology
■ Classification of organisms by shared properties
■ Use binomial genus-differentia nomenclature
○ Galen (130-210 AD)
■ Systematic description of diseases, signs and symptoms
■ In De Febrium Differentia description of fever symptoms he uses the
Aristotelian genus-differentia approach
● Genus-differentia nomenclature
○ Genus-differentia are definitions are key to good ontologies
○ A type of definition where necessary and sufficient conditions are specified that
are composed of two parts
■ Genus: serves as a basis for a new definition; all definitions with the same
genus are considered members of that genus
■ Differentia: the portion of the definition that is not provided by the genus
3
