ExaMode: Building Multi-modal Knowledge Graph for Better Diagnostics
0 likes116 views
The Examode project, supported by the European Union's Horizon 2020 grant, focuses on building a multi-modal knowledge graph to improve medical diagnostics through advanced text analytics and semantic data fusion. It aims to enhance knowledge management of diagnosis-related medical data by linking electronic health records and digital pathology with ontologies and thesauri, facilitating better identification of medical cases. The project runs from January 2019 to December 2022 and involves a consortium dedicated to developing tools for efficient knowledge discovery in healthcare.
ExaMode: Building Multi-modal Knowledge Graph for Better Diagnostics
- 1. Project supported by European Union Horizon 2020 grant agreement 825292 Todor Primov, Sirma AI (Ontotext) Andrey Avramov, Sirma AI (Ontotext) Pavlin Gyurov, Sirma AI (Ontotext) Svetla Boytcheva, Sirma AI (Ontotext) Jul 2 , 2020 | 2:00 PM to 3:00 PM CEST ExaMode: Building Multi-modal Knowledge Graph for Better Diagnostics WEBINAR
- 2. Outline ▪ ExaMode project objectives ▪ Knowledge management of diagnosis-related medical data ▪ Demonstration of services: ▪ Advanced text analytics for semantic data normalization of EHR extracts ▪ Semantic data fusion of extracted results with a referential knowledge graph built from relevant ontologies and thesauri (Mondo Disease Ontology, Disease Ontology, UMLS, SNOMED-CT, etc.) ▪ Visual graph analytics and exploration of semantically normalized cases in the context of the referential knowledge graph ▪ Graph similarity search for the identification of similar medical cases ▪ Discussion 2
- 3. ExaMode - EXtreme-scale Analytics via Multimodal Ontology Discovery & Enhancement https://www.examode.eu/ Duration: Start 01.01.2019, Finish 31.12.2022 Fact sheet Programme(s) H2020-EU.2.1.1. - INDUSTRIAL LEADERSHIP - Leadership in enabling and industrial technologies - Information and Communication Technologies (ICT) Topic(s) ICT-12-2018-2020 - Big Data technologies and extreme-scale analytics Call for proposal H2020-ICT-2018-2 Coordinator: HAUTE ECOLE SPECIALISEE DE SUISSE OCCIDENTALE, Switzerland
- 4. Consortium
- 5. Driven by data, developed for patients 5 Easy and fast, weakly supervised knowledge discovery of exa-scale heterogeneous data, also in highly specific domains. Develop algorithms and tools to link visual content to the associated diagnoses or text and refine the results.
- 6. Data 6 Electronic Health Records (EHR) Digital Pathology Clinical images (Whole Slide Images) Al definitivo si conferma: presenti cellule tumorali maligne che depongono per carcinoma scarsamente differenziato compatibile con carcinoma a piccole cellule del polmone. Su materiale citoincluso presenti frammenti cartilaginei e minuti aggregati neoplastici coerenti con il suddetto reperto. Digital Pathology Diagnostic Reports Cases: Colon cancer Lung cancer Uterine Cervix Coeliac Disease Scientific Publications OPEN DATA
- 7. Driven by data, developed for patients
- 8. What is the key role of ontologies and thesauri in semantic data fusion? How to extract structured data from medical records? Behind the scenes of linking histopathological data and knowledge graphs 8 What are the steps from knowledge discovery and exploration to the medical professional's assistance?
- 10. 10
- 11. How to extract structured data from medical records? 11
- 12. Ontologies, Classifications, Vocabularies • UMLS (Unified Medical Language System) • SNOMED-CT (SNOMED CT catalogue codes for diseases, symptoms and procedures) • MESH (Medical Subject Headings) • ICD-10-CM (International Classification of Diseases, 10th Revision, Clinical Modification) • DOID ( Human Disease Ontology ID) • MONDO (Monarch Disease Ontology) • HPO (Human Phenotype Ontology) • MedDRA (Medical Dictionary for Regulatory Activities Terminology) • Minimal Standard Terminology • etc. 12
- 19. Similarity Search • Predication-based Semantic Indexing (PSI) – Vector Space Model • The similarity plugin allows exploring and searching semantic similarity in RDF resources. • As a user, you may want to solve cases where statistical semantics queries will be highly valuable • Another type of use case is the clustering of patients data into groups by specific patterns.