Extracting body function information using rule-based methods: Highlighting structure and formatting challenges in...

Area of Science:

• Clinical Informatics
• Natural Language Processing
• Biomedical Text Mining

Background:

• Clinical text contains vital information on body function (BF) impairments.
• Identifying BF mentions is crucial for understanding patient health.
• Existing NLP methods struggle with the complex formatting of clinical documents.

Purpose of the Study:

• To identify body function (BF) mentions in a large, heterogeneous clinical text corpus.
• To highlight structural challenges in clinical text that impede NLP techniques.
• To adapt and evaluate a rule-based system for BF mention identification on a new dataset.

Main Methods:

• Utilized a rule-based system initially developed for National Institutes of Health records.
• Adapted the system to a larger, more challenging corpus from the Social Security Administration.
• Focused on identifying body function (BF) mentions within complex clinical text structures.

Main Results:

• Demonstrated the challenges posed by highly formatted clinical text structures.
• Showcased the adaptation of a rule-based system to a different, larger dataset.
• Established a baseline for future improvements in document decomposition and NLP segmentation.

Conclusions:

• Clinical text's structural complexity presents significant challenges for NLP tasks like BF mention identification.
• Adapting existing NLP systems requires careful consideration of data heterogeneity and formatting.
• The developed baseline is foundational for advancing clinical text analysis and document decomposition techniques.