ruleml 2015: ontology reasoning using rules in an ehealth context

ELIS – Multimedia Lab

Dörthe Arndt, Ben De Meester, Pieter Bonte, Jeroen Schaballie,

Jabran Bhatti, Wim Dereuddre, Ruben Verborgh, Femke Ongenae,

Filip De Turck, Rik Van de Walle, and Erik Mannens

Multimedia Lab, Ghent University - iMinds, Belgium

Internet Based Communication Networks and Services , Ghent University - iMinds, Belgium

Televic Healthcare - Izegem, Belgium

RuleML 2015, Berlin, August 05, 2015

Ontology Reasoning using Rules

in an eHealth Context


Ontology Reasoning using Rules in an eHealth Context

Outline

Business Case

Technological Challenges

Rule Based Solution

Results

Importance and Impact



Adaptable context aware nurse call system for

Why do they want that?

• Less distraction for nurses, if they are busy they don’t get called

• Less walking distances for nurses and doctors

• No time loss by assigning calls to persons without the necessary competences

• Trust between personell and patients

• Adaptivity to the requirements of any hospital

→More efficient organization of hospitals

Business Case



Outline

Business Case


Rule Based Solution

Results





• Scalability cope with data sets ranging from 1000 to 100 000 relevant triples

• Semantics be able to draw conclusions based on the information it is aware of

• Functional complexity implement deterministic decision trees with varying complexities

• Configuration have the ability to change these decision trees at configuration time

• Real-time return a response within 5 seconds to any given event



Why rule based solution?

Classical (Java, C++, …)

OWL DL + SPARQL

Rule based

Scalability

Semantics

Functional Complexity

Configuration

Real Time



Outline

Business Case


Rule Based Solution

Results




Available data



Available data

Our ontology (ACCIO ontology) is filled with:



Available data


• Location of personnel and patients



Available data



• Current task of care givers (what is he/she doing?)



Available data




• Competences of staff members



Available data





• Special needs of patients



Available data






• Relationship of nurses and patients



Available data







• (Possible) Reasons for calls



Available data







• (Possible) Reasons for calls

And much more…



• For reasoning we used the EYE reasoner

• We used Notation3 Logic to express

• OWL RL rules

• Decision Trees

Rules



OWL RL in N3

As an example we take subClassOf:

{?C rdfs:subClassOf ?D. ?x a ?C.} => {?x a ?D.}.



OWL RL in N3

Knowledge:

:Call rdfs:subclassOf :Task. :call1 a :Call.





OWL RL in N3

Knowledge:



We get:




OWL RL in N3

Knowledge:



We get:

:call1 a :Task.




Decision Tree

{

?c rdf : type : Call .

?c : hasStatus : Active .

?c : hasReason [rdf: type : CareReason ].

?p rdf : type : Person .

?p : hasStatus : Busy .

?p : hasRole [rdf: type : StaffMember ].

?p : hasCompetence [

rdf: type :

AnswerCareCallCompetence

].

}

=>

{

(?p ?c) : assigned 100.

}.

{



?c : madeAtLocation ?loc.


?p : hasStatus : Free .

?p : closeTo ?loc.

}

=>

{


}.

Priority

0

200

100



Decision Tree

{








rdf: type :


].

}

=>

{


}.

{



?c : madeAtLocation ?loc.


?p : hasStatus : Free .

?p : closeTo ?loc.

}

=>

{


}.

Priority

0

100

200 Priorities can be changed easily



Decision Tree

{








rdf: type :


].

?c :madeAtLocation ?loc.

?p :closeTo ?loc.

}

=>

{


}.

New triples can be added easily



Outline

Business Case


Rule Based Solution

Results




Testscenario

1. A patient launches a call (assign nurse and update call status) 2. The assigned nurse indicates that she is busy (assign other nurse) 3. The newly assigned nurse accepts the call task (update call

status) 4. The nurse moves to the corridor (update location) 5. The nurse arrives at the patients’ room (update location, turn on

lights and update nurse status) 6. The nurse logs in to the room’s terminal (update status call and

nurse, open lockers) 7. The nurse logs out again (update status call and nurse, close

lockers) 8. The nurse leaves the room (update location and call status and

turn off lights)



Results: 1 ward



Results: 10 wards



Outline

Business Case


Rule Based Solution

Results




We learned:

• Applying rule based reasoning instead of OWL DL & SPARQL makes a difference

• First results are promising

• For small data sets we meet the requirements

• Reasoning times are stable

• Decision trees are easy to handle via rules

→ Further improvements will lead to faster implementations


Rule based reasoning can be used in future products of televic healthcare

ruleml 2015: ontology reasoning using rules in an ehealth context

Science