T-staging pulmonary oncology from radiological reports using natural language processing: translating into a...

Area of Science:

• Medical informatics
• Natural Language Processing (NLP)
• Oncology

Background:

• Accurate medical data structuring is crucial for oncological staging, patient treatment, and population surveillance.
• A Dutch NLP algorithm was developed to quantify pulmonary tumor T-stage from free-text radiological reports.
• This algorithm was translated and validated for use with English reports.

Purpose of the Study:

• To develop and validate a natural language processing (NLP) tool for automated T-stage extraction from English radiological reports.
• To assess the accuracy of the NLP algorithm in classifying T-stage for lung cancer staging.
• To compare the performance of the English NLP tool with its Dutch counterpart.

Main Methods:

• A rule-based NLP algorithm was trained on 200 English free-text radiological reports.
• The algorithm was validated on 225 English diagnostic computed tomography (CT) reports for lung cancer staging.
• Performance was evaluated by comparing automated T-stage extraction against manual staging, with a graphical user interface for visualization.

Main Results:

• The T-stage classifier achieved an accuracy of 0.89 in the validation set.
• Accuracy was 0.84 for T-substages and 0.76 for tumor size alone.
• Performance was comparable to the Dutch version, with most errors stemming from algorithmic ambiguity.

Conclusions:

• NLP is effective for staging lung cancer from radiological reports across different languages.
• A hybrid approach incorporating machine learning can further enhance NLP algorithm performance.