Evolving Medical Ontologies Based on Causal Inference
Autor: | Hengyi Hu, Larry Kerschberg |
---|---|
Rok vydání: | 2018 |
Předmět: |
Hierarchy
business.industry Computer science Bayesian network 02 engineering and technology Ontology (information science) Causality Data science Terminology Constraint (information theory) Analytics 020204 information systems Causal inference 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing business |
Zdroj: | ASONAM |
Popis: | Causal inference and analytics plays a critical role in public health and disease prevention. Through mining of large patient datasets, it is possible to identify opportunities for intervention and to determine the effectiveness of treatment. There are currently many methods to analyze and learn causal relationships in large patient datasets, as well as specific causal studies in epidemiology that define specific relationships among symptoms and treatments. This paper introduces a novel methodology to utilize causal knowledge to extend and improve a standard hierarchical medical ontology. First, we will obtain the hierarchical structure of the patient symptom variables based on the Medical Dictionary for Regulatory Activities Terminology (MedDRA). Then, we will learn a Causal Bayesian Network (CBN) using Max-Min Hill-Climbing (MMHC), a hybrid constraint and score-based learning algorithm, on the pre-existing National Institutes of Mental Health (NIMH) study on Sequenced Treatment Alternatives to Relieve Depression (STAR*D) patient dataset. Finally, we will use the causal links discovered in the CBN to evolve the ontology and its hierarchy. |
Databáze: | OpenAIRE |
Externí odkaz: |