Using natural language processing to automatically classify written self-reported narratives by patients with...

Area of Science:

• Computational linguistics
• Medical informatics
• Clinical neurology

Background:

• Headache diagnosis relies heavily on physician interpretation of patient-reported symptoms.
• Natural Language Processing (NLP) offers computational methods to analyze and structure linguistic data.
• This study explores NLP and machine learning (ML) for analyzing patient narratives in headache disorders.

Purpose of the Study:

• To investigate the potential of NLP for analyzing self-reported headache narratives.
• To apply ML algorithms for automatic classification of headache types (migraine vs. cluster headache).
• To assess the feasibility of information extraction from clinical text data.

Main Methods:

• Collected self-reported narratives from 121 patients (81 migraine, 40 cluster headache).
• Applied NLP for lexical, semantic, and thematic analysis of patient texts.
• Utilized ML algorithms (logistic regression, support vector machine) for classification.

Main Results:

• Identified distinct keywords for cluster headache (e.g., 'eye', 'pain') and migraine (e.g., 'headache', 'stress', 'nausea').
• Both patient groups exhibited predominantly negative sentiment in their narratives.
• ML models, particularly logistic regression and SVM, achieved high F1-scores (0.82-0.86) for classifying cluster headache descriptions.

Conclusions:

• NLP successfully identified lexical differences between migraine and cluster headache narratives, aligning with clinical expertise.
• ML algorithms demonstrate strong potential for accurately classifying headache types from patient-written text.
• NLP is a valuable tool for clinical information extraction and analyzing patient-generated data in headache disorders.