Towards generating a patient's timeline: extracting temporal relationships from clinical notes.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• Clinical records contain complex patient histories in both coded and free-text fields.
• Extracting temporal relations between events is crucial for building patient timelines automatically.

Purpose of the Study:

• To develop and evaluate a novel hybrid system for extracting temporal relations from clinical notes.
• To improve the automatic construction of patient event timelines.

Main Methods:

• A hybrid system combining machine learning with graph-based inference was designed.
• Temporal relation extraction components were developed for different relation types.
• A temporal graph based on parse tree dependencies and pattern clues was utilized.
• Sentence generalization was employed to discover hidden patterns.

Main Results:

• The hybrid system achieved an F-measure of 0.63 (precision 0.76, recall 0.54) on the 2012 i2b2 temporal relation extraction task.
• The system demonstrated high precision and a competitive F-measure among participating teams.

Conclusions:

• The proposed hybrid system effectively extracts temporal relations from clinical text.
• This method advances the automatic construction of patient timelines from electronic health records.