An ontological approach for the exploitation of clinical data Ariane Assélé Kama 1 ,Rémy Choquet 1 , Giovanni Mels 2 , Christel Daniel 1,3 , Jean Charlet 1,3 , Marie-Christine Jaulent 1 1 INSERM, UMR_S 872, Eq. 20, Université Pierre et Marie Curie, Paris, France; 2 AGFA HealthCare NV, Moutstraat 100, 9000 Gent, Belgium; 3 APHP, Assistance Publique des Hôpitaux de Paris, Paris, France; MEDINFO – Copenhagen, 22 th of August 2013 Oral Presentation Presenter: Ariane Assélé Kama 1 Assises GDR I3 – Strasbourg 01/07/2010
DebugIT Detecting and Eliminating Bacteria Using Information Technology European project FP7 (grant #217139) with 14 partners. q Aggregate data stored in European hospitals q Build a unified system q Infectious disease control q Antimicrobial resistances Disclaimer: this presentation reflects solely the views of the authors and no guarantee or warranty is given that it is fit for any particular purpose. The European Commission, Directorate General Information Society and Media, Brussels, is not liable for any use that may be made of the information contained therein. 2 Assises GDR I3 – Strasbourg 01/07/2010 Slide 2 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Hospital Information Systems Introduction Analysis Space ETL Process Clinical Databases Decision making Extract – Transform – Load Consolidate, archive Data analysis Diagnostic or Therapeutic decisions Large data streams Data mining Clinical or Epidemiological research 3 Assises GDR I3 – Strasbourg 01/07/2010 Slide 3 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Hospital Information Systems Why we need to use semantics resources ? Data Sources ID Bacterium Antibiotic Result Value ID Bactérie Antibiotique R V 1 E. Coli Amoxicilline Resistant 25 10 Escherichia.Coli Cefpirome R 25 20 Hafnia alvei Quinolones S 28 2 Escherichia coli Cefoperazone Sensitive 28 30 Pseudomonas aeruginosa Ofloxacine I 10 3 Pseudomonas aeruginosa Ofloxacine Intermediate 10 Question What are the results of susceptibility testing of E. Coli resistant to B-lacatam? Problem � Lack of knowledge : Amoxicillin is a B-lactam � E. Coli and Escherichia coli refer to the same bacteria � Fields “bacterium” & “bactérie” refer to the same “bacteria” concept Hypothesis Using an ontology to enrich clinical data exploitation Objective To exploit domain knowledge throughout ontology, as a user oriented view, to query clinical data, building a SPARQL Endpoint. 4 Assises GDR I3 – Strasbourg 01/07/2010 Slide 4 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Material & Methods SPARQL Endpoint Building Process (a) Traditionnal multidimentionnal datawarehouse D2R Server / Joseki 2 - Mapping File: (concepts) ó (tables/colums) - RDF Graphe (fichier .n3) - Data mapping SPARQL Endpoint 3 - Clinical Data access - SPARQL queries 1 Ressources - Clinical Database (HEGP) - DCO 7 : Ontologie (DebugIT) (b) SPARQL Endpoint building approach 5 Assises GDR I3 – Strasbourg 01/07/2010 Slide 5 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
SPARQL Endpoint building Process Database mapping file DebugIT Core Ontology (DCO) Database Information Model 6 Assises GDR I3 – Strasbourg 01/07/2010 Slide 6 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
SPARQL Endpoint building Process Database mapping to DCO Ontology Data Concepts 7 Assises GDR I3 – Strasbourg 01/07/2010 Slide 7 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
SPARQL Query Example Illustration of a SPARQL Query building PREFIX biotop : <http://purl.org/biotop/1.0/biotop.owl#> PREFIX dco: <http://www.debugit.eu/ontology/1.0/dco.owl#> PREFIX inserm: <http://debugit1.spim.jussieu.fr/resource/biotop.owl#> SELECT DISTINCT * WHERE { Mapping file between data and ontology GRAPH <http://debugit.eu/inserm-map.n3> { ?antibiotic1 a dco:BetalactamAntibiotic. ?bacteria a biotop:SpeciesEcherichiaColiRegion. } We get the corresponding URIs of GRAPH <http://debugit1.spim.jussieu.fr/resource> the B-Lactam antibiotics and E. coli { bacteria ?susceptibility a inserm:ResultAntibiogram; inserm:antibiogram_ID ?antibiogram; From the mapping file, specifications inserm:antibiotic_tested ?antibiotic1; data from the database are recovered inserm:antibiotic_RESULT ?r1. ?antibiogram a inserm:antibiogram; inserm:bacteria_analyzed ?bacteria. } } 8 Assises GDR I3 – Strasbourg 01/07/2010 Slide 8 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Results SPARQL Query running Initial data � 238 623 antibiogram results � Performed on 61 antibiotics, including 22 Beta-Lactam � 165 bacteria SPARQL Query result D2R Server 4 075 results of susceptibility of resistance of E. Coli to beta-lacatam in 24.6379 s Joseki : using ontology concept 4 075 results of susceptibility of resistance of E. Coli to beta-lacatam in 10.085 s 9 Assises GDR I3 – Strasbourg 01/07/2010 Slide 9 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Results Comparison with the HEGP microbiologists reports E. Coli vs CEFIXIM: sensitive at hegp vs debugit (%) 20 20 20 20 20 20 2007 01 02 03 04 05 06 HEGP 73 85 90 91 91 91 89% (Internal report) DebugIT 73 86 89 91 92 91 91% (Sparql Endpoint) Graphic rate of E. Coli sensitivity to Cefixim at HEGP Total 12 12 28 27 28 28 2727 hospital over a period of 6 years. 44 44 53 80 38 50 ddo:Sensitive 91 20 25 25 26 25 2479 1 74 53 25 07 96 Rate sensitivity of E. coli to Cefixim at HEGP over a period of 6 years. 10 Assises GDR I3 – Strasbourg 01/07/2010 Slide 10 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Limits and Future Work Conclusion Using semantic resource to exploit clinical data Ontologies are used to enrich the data analysis process q We run fewer queries using ontology and retrieve all data q Semantic relations q “is-a” hierarchy was used in this study q All existing semantic relations could be used (e.g. equivalence) Database mapping to ontology Mapping manually done q Link an instance to a concept q Perspectives Build a data definition ontology from the database information model q Define mapping rules between the domain ontology (DCO) an the data definition q 11 Assises GDR I3 – Strasbourg 01/07/2010 Slide 11 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
The DebugIT European Projet Conclusion The Semantic Interoperability Platform 12 Assises GDR I3 – Strasbourg 01/07/2010 Slide 12 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Acknowledgments DebugIT, EU-IST-FP7-217139 DebugIT partners EAGL, SNF-325230-120758 Agfa Healthcare (Belgium) Empirica (Germany) Agfa Healthcare Gama Sofia Ltd (Bulgaria) • Giovanni Mels INSERM (France) • Hans Cools IZIP (Czech Republic) Linköping University (Sweden) INSERM EQ20 UMRS 872 TEILAM (Greece) University College London (UK) HUG (Switzerland) Freiburg University (Germany) Geneva University (Switzerland) Averbis (Germany) MDA (Czech Republic) HES-SO (Switzerland) 13 Assises GDR I3 – Strasbourg 01/07/2010 Slide 13 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Thank you for your attention ariane.asselekama@gmail.com 14 Assises GDR I3 – Strasbourg 01/07/2010 Slide 14 MEDINFO Copenhague 22 th of August 2013 Presented by Ariane Assélé Kama
Recommend
More recommend
