Design of an automatic coding algorithm for a multi-axial classification in pathology

Area of Science:

• Medical Informatics
• Natural Language Processing
• Pathology Reporting

Background:

• Automatic coding of pathology reports is crucial for data extraction and analysis.
• Existing methods may struggle with the complexity and variability of medical terminology.
• A multi-axial codification system (ADICAP) offers a structured approach.

Purpose of the Study:

• To describe and evaluate the ADICAP algorithm for automatic pathology report coding.
• To assess the algorithm's ability to extract significant words and their statistical relationships.
• To identify areas for improvement, particularly in organ-specific coding.

Main Methods:

• The ADICAP algorithm was developed to extract significant words and expressions from pathology reports.
• Statistical relationships between extracted terms and multi-axial codification modalities were recorded.
• Various weighting functions were evaluated to optimize performance.

Main Results:

• The algorithm achieved high accuracy in identifying correct modalities within the top 5 candidates for most axes.
• Performance was notably lower for the organ axis, indicating a specific challenge.
• Evaluation of different weighting functions informed optimal algorithm settings.

Conclusions:

• The ADICAP algorithm demonstrates significant potential for automated pathology report coding.
• Further development is required to enhance accuracy for the organ axis.
• Future work may involve a two-stage algorithm to improve overall performance.