Automatic classification of radiological reports for clinical care

Area of Science:

• Radiology
• Medical Informatics
• Natural Language Processing

Background:

• Radiological reports contain valuable free-text clinical narratives.
• Automated analysis is needed to aggregate and utilize this information effectively.
• Chest computed tomography (CT) reports are a key data source.

Purpose of the Study:

• To develop and evaluate a novel hierarchical classification system for chest CT reports.
• To classify reports based on a schema developed by clinical radiologists.
• To leverage machine learning for automated report analysis.

Main Methods:

• Utilized a training dataset of radiologist-annotated chest CT reports.
• Developed a hierarchical classification system using a cascade of machine learning classifiers.
• Employed syntactic and semantic features for classification.

Main Results:

• A novel hierarchical classification system for chest CT reports was developed.
• The system was trained on radiologist-annotated data and textual evidence.
• Experimental evaluation demonstrated the system's effectiveness.

Conclusions:

• Automated classification of chest CT reports is feasible and valuable.
• Hierarchical machine learning approaches can effectively analyze clinical narratives.
• The developed system enhances the utility of radiological reports for care and research.