A Preliminary Study of Clinical Concept Detection Using Syntactic Relations.

Area of Science:

• Clinical Natural Language Processing
• Biomedical Informatics
• Computational Linguistics

Background:

• Concept detection is crucial for clinical Natural Language Processing (NLP).
• Clinical concepts are complex, with diverse expressions and synonyms.
• Challenges arise from discontinuous and permuted phrase occurrences in clinical text.

Purpose of the Study:

• To investigate an approach for improving clinical concept detection.
• To leverage syntactic information to overcome challenges in identifying complex clinical concepts.
• To enhance the recall of concept detection in clinical notes.

Main Methods:

• Mining syntactic patterns from continuous, non-permuted concept synonyms.
• Utilizing mined syntactic patterns to detect discontinuous and permuted concept phrases.
• Experimenting on 790 de-identified clinical notes.

Main Results:

• The proposed approach shows potential for boosting concept detection recall.
• Preliminary experiments indicate improved performance in identifying clinical concepts.
• Challenges and limitations were identified during the analysis.

Conclusions:

• Exploiting syntactic information is a promising strategy for enhancing clinical concept detection.
• The approach offers a potential solution for handling complex and varied clinical concept expressions.
• Further research is needed to address identified limitations and fully realize the approach's potential.