An Empirical Method of Automatic Pattern Extraction for Clinical Text Classification.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Clinical Text Mining

Background:

• Clinical text classification is crucial in medical text processing.
• Existing methods often rely on manual pattern identification, limiting efficiency.
• Pattern-based approaches generally outperform other methods in this domain.

Purpose of the Study:

• To develop and evaluate a novel, automated pattern extraction algorithm for clinical text classification.
• To address the limitations of human intervention in traditional pattern-based methods.
• To improve the accuracy and applicability of clinical text classification.

Main Methods:

• Developed an algorithm to automatically identify candidate concepts and their context windows in clinical text.
• Transformed context windows into patterns for classification.
• Evaluated the algorithm on clinical guidelines for Hypertension, Rhinosinusitis, and Asthma.

Main Results:

• The algorithm successfully extracted 21 patterns.
• Achieved high accuracy in classifying recommendation and non-recommendation sentences: 84.53% for Hypertension, 80.03% for Rhinosinusitis, and 84.62% for Asthma.
• Demonstrated the algorithm's effectiveness on diverse clinical guidelines.

Conclusions:

• The proposed automated pattern extraction algorithm offers significant benefits for clinical text classification.
• The method is applicable and efficient for processing clinical textual resources.
• This approach reduces the need for manual intervention, enhancing the utility of pattern-based methods.