THBP

THBP maps anatomical entities mentioned in free-text medical documents, such as discharge summaries, to human body parts using a Wikipedia-based scoring algorithm and the Tree of Human Body Parts ontology to support medical text mining.

Key Features:

• Wikipedia-Based Scoring Algorithm: Uses Wikipedia's medical content to evaluate and score candidate mappings between anatomical entities and body parts.
• Named Entity Normalization (NEN): Standardizes variants of anatomical terminology to canonical forms to enable consistent identification and mapping.
• THBP Ontology: Tree of Human Body Parts ontology constructed with anatomical expert input and aligned with the Unified Medical Language Systems (UMLS) to categorize core anatomical parts.

Scientific Applications:

• Medical Text Mining: Enhances extraction of body-part associations from clinical narratives for research and analysis.
• Clinical Treatments: Provides structured mappings of anatomical entities to support clinical decision-making, diagnosis, and treatment planning.
• Evaluation and Performance: In an evaluation on 50 discharge summaries containing 2,224 anatomical entities, the THBP-based algorithm achieved an F1-measure of 86.67% versus a baseline of 70.20%.

Methodology:

Two-step computational process: (1) entity recognition and named entity normalization to canonical anatomical terms; (2) mapping normalized entities to the THBP ontology using a Wikipedia-based scoring algorithm that assigns scores reflecting mapping likelihood.