THBP

THBP (Tree of Human Body Parts) is a software tool to map anatomical entities mentioned in medical narratives, such as discharge summaries, to a standardized ontology of human body parts. By leveraging information from Wikipedia and the Unified Medical Language Systems (UMLS), the tool aims to improve the accuracy and efficiency of medical text mining applications.

Key features of THBP include:

1. A Wikipedia-based scoring algorithm that utilizes the comprehensive medical knowledge available on the platform to map anatomical entities to their corresponding body parts accurately.

2. A named entity normalization (NEN) method that standardizes the anatomical entities mentioned in the medical narratives, making the mapping process more efficient.

3. A custom-built ontology of core anatomical parts, created with the help of anatomical experts and UMLS, to ensure the mapping system is effective for clinical treatments.

The performance of THBP was evaluated using a gold standard dataset derived from 50 discharge summaries containing 2,224 anatomical entities. The tool achieved an F1 measure of 86.67%, significantly outperforming the baseline system, which had an F1 measure of 70.20%.