MVilla IUI 2012 Lisbon

1,465 views

Published on

A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition

Published in: Technology, Travel, Education
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
1,465
On SlideShare
0
From Embeds
0
Number of Embeds
471
Actions
Shares
0
Downloads
9
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

MVilla IUI 2012 Lisbon

  1. 1. A Learning Support Tool with Clinical Cases Based onConcept Maps and Medical Entity Recognition Manuel de la Villa1, Fernando Aparicio2, Manuel J. Maña1, Manuel de Buenaga2 1Universidad de Huelva, 2Universidad Europea de Madrid Presenting Prof. Mr. Manuel de la Villa manuel.villa@dti.uhu.es http://www.uhu.es/manuel.villa
  2. 2. Index  The problem. An Use Case.  Related work. - Biomedical Ontologies - Concept map and Mind map - Graph-based Interfaces based on Ontologies  A rough prototype as a “proof of concept”.  Evaluation  Conclusions and future works.A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 2
  3. 3. The scenario   The use of intelligent systems in higher education is incresingly used as strategy to improve learning and teaching processes.   The student reads new concepts, he needs more   Case-based learning, based on constructivist information to understand them. learning theories, is very practical in Medical education.   HOW???   Making the internet sources available to students may not be sufficient to promote   A free search? their learning… let’s see an example.   One for every term??A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 3
  4. 4. The problem (I)  Physicians in the early stages of learning face several drawbacks among [Luo & Tang 2008]: -  Lack of experience and domain knowledge to perform a proper search -  Lack of awareness about the medical terminology found Oughhh!A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 4
  5. 5. The problem (and II)   Free search???   User have problems to define their information needs in a query string [Jansen, Spink & Koshman, 2007].   Queries contain less than three terms (75,2%) and the majority of queries contain one (18,5%), two (32,2%)   But also when the user initiates a search not really know what can be useful and, therefore, it is difficult to specify the features of the elements of potentially useful information [Belkin, 2000]. Search engines usually return thousands of documents recovered, leading to inadequate results, with no semantic connection with the query and little to do with the users needs.A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 5
  6. 6. Our proposal The design of a support tool for Clinical Case-based learning that… Freebase … helps clinicians identifyNLP access the meaning of medical and NCBO Open MQL Topics Biomedical concepts and … Module Annotator Concepts table … allow the teacher Search Module … display concept UMLS Graph Module to define the paths maps automatically of access to drawn from knowledge Concepts map Freebase information Freebase Medlineplus sources. avoiding dispersion in the search andA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 6
  7. 7. Related work: Biomedical Ontologies   May include a wide range of medical concepts, basic information such as the type or class they belong to and how they are related (e.g. symptom / disease / treatment). Is-a-symptom-of Is-treated-with Jaundice Hepatitis Adefovir   Increasingly used to tackle concept recognition and annotation tasks in biomedical research.   Some examples of ontologies are: -  GO (Gene Ontology), MeSH (Medical Subject Headings), FMA (Foundational Model of Anatomy), GALEN, UMLS (Unified Medical Language System), SNOMED-CT (Systematized Nomenclature of Medicine - Clinical Terms), etc. We decide to use MedlinePlus (Health Topics), Freebase and UMLS mainly due to the ease of open information access through web services and XML filesA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 7
  8. 8. Ontologies usedUMLS Metathesaurus  UMLS (Unified Medical Language o Remote access with UTS Web System), developed by the National services API. Library of Medicine (NLM) of USA. o Source: MDR, The Medical o Metathesaurus Dictionary for Regulatory Activities o  Concept (MedDRA), developed by ICH, owned by IFPMA. o  CUI (Concept Unique Identifier) o Translations: Czech, Dutch, o  Semantic Type(s) French, German, Italian, Japanese, o  Definition (if provided) Portuguese and Spanish. o  Atoms o  Contexts o  Concept Relations
  9. 9. Ontologies used Metaweb Freebase •  Freebase is a large collaborative knowledge base consisting of metadata composed mainly by data integration processes and by its community members. •  Domain independent nature: possibilities of applying results to other disciplines. •  The information can be accesed through an API, MQL (Metaweb Query Language), ACRE (an own platform to host applications) o RDF. Our MQL Query for Concepts Map: http://api.freebase.com/api/service/mqlread?query= {"query":”[{"type":"/medicine/disease", "name":""+search_string+"","/common/topic/article":{"guid":null,"limit":1,"optional":true}, "/common/topic/image": {"id":null,"limit":1,"optional":true},"symptoms":[],"treatments":[], "/medicine/disease/notable_people_with_this_condition": [],"/medicine/disease/risk_factors": [], "/medicine/disease/causes": [],"/medicine/disease/prevention_factors": []}]}A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 9
  10. 10. Ontologies usedMetaweb Freebase Ontology fragment for biomedical domain in Freebase
  11. 11. Concept map and Mindmap approaches.   Widely applied in educational activities   2-dimensional graphics used to represent knowledge comprised of nodes (representing concepts) connected by direct arcs (representing relationships)A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 11
  12. 12. Related work: Concept map and Mindmap approaches.  Advantages: -  Graphic presentation of knowledge enables quickly evaluation for experts -  In medical studies: -  [Daley & Torre, 2010] Concept mapping in medical and healthcare learning: -  Promotes learning, provides additional resources, provides feedback to students and conducts assessment -  [D’Antoni et al., 2009] Mind maps are very useful in medical education. -  Problems: many topics to be covered in medicine, fair amount of time to design them  Knowledge visualization, an emerging field.  Similarities between ontologies and concept maps.A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 12
  13. 13. Our metaphor? A graph (Concept Map)   Concept Map extracts and displays only the information needed to determine a diagnosis of a disease in a medical case.A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 13
  14. 14. Graph-based Interfaces based on Ontologies Information retrieval   Visual Concept Explorer: an automatic concept map generator with knowledge from medical ontologies and thesauri.A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 14
  15. 15. Graph-based Interfaces based on Ontologies Visual dictionaries   Based on a Thesaurus (Wordnet™) Visual Thesaurus Snappy WordsA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 15
  16. 16. Graph-based Interfaces based on Ontologies Search engines Wikimindmap builds a mental map from the information you find on a concept in the Wikipedia. It could be considered as a dynamically and automatically generated interface to browse Wikipedia.A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 16
  17. 17. Graph-based Interfaces based on Ontologies Search engines Yahoo Correlator extracts and organizes Google Wonder Wheel shows related search information from text, and searches for related terms to the current searched query and thus names, concepts, places, and events to your query. enable you to explore relevant search terms.A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 17
  18. 18. Graph-based interfaces based on ontologiesSemrep   SemViz (Semantic Abstraction Summarization [Rindflesh, Fiszman and Kilicoglu, 2004])   Takes as input a list of semantic predications produced by UMLS SemRep, from a set of documents on a specified disorder topic. The output is a conceptual condensate (a concept map in graphic format) containing only those predications that represent key information in the input documents.A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 18
  19. 19. Computer tool description http://orion.esp.uem.es:8080/MedicalFaceV2/A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 19
  20. 20. Computer tool description Freebase NLP NCBO Open MQL Topics Biomedical Module Annotator Concepts table Search Module Graph Module UMLS Concepts map Freebase Freebase MedlineplusA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 20
  21. 21. The system working… http://youtu.be/Dp9flQpvJdE http://www.medicalminer.org/MedicalFaceV2/ http://www.uhu.es/manuel.villa/viewmed http://sciencecases.lib.buffalo.edu/cs/files/A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition stroke.pdf 21
  22. 22. User evaluation  User oriented evaluation -  Users: 60 second-year medical degree students from the School of Biomedical Sciences at the Universidad Europea de Madrid, divided into 2 groups. -  Objectives: To measure the influence of the system when student make a test, besides usability and learning support provided. -  Technique: -  Exam with 10 multiple choice questions about a selected case study -  34 self-perception Likert questionnaires for system users.  Measure the differences between the results of the activity carried out in two ways: Mitral  regurgitation: a.-­  Is  the  less  common  valvulopathy  in  the  general  population   -  With the system developed b.  -­  Has  no  relation  with  the  cardiac  problem  presented  by  our  patient -  With free Internet access c.  -­  May  justify  the  mitral  regurgitation d.  -­  Has  a  higher  prevalence  in  women  than  in  men Test  question  exampleA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 22
  23. 23. Results user evaluation   Slightly better results for students who employed the tool (78.53% correct answers) than students who used unrestricted searches (76.92% correct answers). No statistically significant. Learning perception questions • O ver 58% believe that the tool has helped them to extract relevant information about the case study (LQ1), and • more than 60% believe that the tool has helped them by reducing the time needed to understand the case study (LQ2). Students  learning  self-­perceptionA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 23
  24. 24. Results user evaluation Usability questions: • the tool interface is nice (UQ1), •  it is easy to find the information required (UQ2), • they feel comfortable using the tool (UQ3), • the speed is reasonable (UQ4) and Students  usability  self-­perception • it is easy to use (UQ5).A Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 24
  25. 25. Systematic evaluation   measure the ability of the tool to provide medical concepts in the graph, in relation to the original concepts annotated in the source document (as recall in information retrieval)   measure novelty, the tool’s ability to discover and show us new relevant information related with the source document. CrFreebase ∑corpus Ca SnomedCT + CrFreebase Novelty( corpus) = N # documentsA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 25 €
  26. 26. Conclusions.   interfaces that simplify finding and comprehension of information are needed.   we have presented a tool that represent biomedical knowledge resources in a human and machine usable way (as ontologies and concept maps)   the knowledge acquired through an active role is better fixed in their minds and longer term.   advantage for teachers: it allows pre- selection of the knowledge sources accessible to students.   The students’ perception is good or very good in both usability questions and those related to the assistance providedA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 26
  27. 27. Future work.   Focus our efforts on enhancing all the available features in the tool: -  usability of the interface, -  expansion and improvement of the annotation process and -  enrichment of the information and concept mapping.   Expand the user experience evaluation, to measure the tool’s capacity to support teachers in active learning methodologiesA Learning Support Tool with Clinical Cases Based on Concept Maps and Medical Entity Recognition 27
  28. 28. Muito ObrigadoA Learning Support Tool with Clinical Cases Based onConcept Maps and Medical Entity Recognition Manuel de la Villa1, Fernando Aparicio2, Manuel J. Maña1, Manuel de Buenaga2 1Universidad de Huelva, 2Universidad Europea de Madrid Presenting Prof. Mr. Manuel de la Villa manuel.villa@dti.uhu.es http://www.uhu.es/manuel.villa

×