Automated Classification of Pathology Reports.

Area of Science:

• Computational pathology
• Natural Language Processing (NLP) for medical documents
• Oncology informatics

Background:

• Pathology report analysis is crucial for cancer diagnosis and treatment.
• Manual classification of tumor topography and morphology is time-consuming and prone to errors.
• Standardized coding systems like ICD-O are essential for data consistency.

Purpose of the Study:

• To develop an automated classifier for pathology reports.
• To infer tumor topography and morphology classes using ICD-O codes.
• To establish a baseline for future research in medical document classification.

Main Methods:

• Utilized Naive Bayes classifiers for automated classification.
• Trained and validated models on a large dataset of 94,980 patient reports from A.C. Camargo Cancer Center.
• Employed the F1-score metric for performance evaluation.

Main Results:

• Achieved high performance with F1-scores exceeding 74% for the topographic group.
• Reached F1-scores above 61% for the morphologic group.
• Demonstrated the effectiveness of automated classification for medical reports.

Conclusions:

• The developed automated classifier offers a successful baseline for classifying medical documents, particularly in Portuguese.
• This approach can significantly improve efficiency and accuracy in pathology report analysis.
• The methodology holds potential for application in other domains and languages.