DiseaSE: A biomedical text analytics system for disease symptom extraction and characterization

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Data Mining

Background:

• The rapid growth of biomedical literature presents challenges in extracting valuable information.
• Unstructured text data in medicine often remains underutilized.

Purpose of the Study:

• To develop a biomedical text analytics system, DiseaSE (Disease Symptom Extraction), for identifying and extracting disease symptoms and their associations.
• To convert unstructured biomedical text into structured information components for analysis.

Main Methods:

• Utilized Natural Language Processing (NLP) and information extraction techniques.
• Processed text documents into semantic triples.
• Employed TextRank and other ranking algorithms to identify disease symptoms.
• Evaluated the system on eight diseases: dengue, malaria, cholera, diarrhoea, influenza, meningitis, leishmaniasis, and kala-azar.

Main Results:

• DiseaSE successfully identified disease symptoms and their associations from PubMed documents.
• The system discovered novel symptoms not cataloged by authoritative sources like CDC, WHO, and NHS.
• Generated graph-theoretic analysis and visualization for disease characterization.

Conclusions:

• The DiseaSE system effectively extracts critical disease-symptom information from biomedical literature.
• Identified symptoms and associations can contribute to creating comprehensive biomedical knowledgebases and ontologies.
• The system supports the development of advanced e-health and disease surveillance systems.