DEEPEN: A negation detection system for clinical text incorporating dependency relation into NegEx

Area of Science:

• Clinical informatics
• Natural Language Processing (NLP)
• Biomedical data mining

Background:

• Electronic Health Records (EHRs) contain valuable patient information in free text.
• Natural Language Processing (NLP) is crucial for extracting clinical information from EHRs.
• Accurate negation detection is vital for correct interpretation of clinical concepts.

Purpose of the Study:

• To develop a novel negation detection algorithm (DEEPEN) to address limitations of existing methods like NegEx.
• To improve the accuracy of identifying negated clinical concepts in complex sentences within EHRs.
• To reduce false positives in negation detection, leading to more reliable clinical data analysis.

Main Methods:

• Developed DEEPEN, a negation algorithm incorporating Stanford dependency parsing.
• Utilized dependency relationships between negation words and clinical concepts.
• Trained and tested the algorithm on EHR data from Indiana University (IU).
• Evaluated generalizability using the Mayo Clinic dataset.

Main Results:

• DEEPEN significantly reduced false positives in negation assignment compared to NegEx.
• The algorithm demonstrated improved accuracy in identifying negated clinical findings.
• Evaluation on diverse datasets confirmed DEEPEN's effectiveness and generalizability.

Conclusions:

• DEEPEN enhances negation detection in clinical NLP by leveraging sentence dependency structures.
• The algorithm improves the accuracy of patient cohort identification from EHRs.
• This advancement contributes to more precise clinical decision-making and research.