Comparison of a Natural Language Processing Model and Large Language Models for Extracting Incidental Lung Nodule Data

Key Points

• To compare the effectiveness of a task-specific natural language processing model with general-purpose large language models in extracting incidental lung nodule data.
• Compared a task-specific natural language processing model with seven large language models.
• Analyzed unstructured chest CT radiology reports for incidental lung nodule data extraction.
• Measured accuracy and reliability of data extracted from reports.
• The natural language processing model outperformed the large language models in accuracy.
• Higher reliability was noted in the task-specific model for data extraction.
• The findings indicate superior performance for clinical data interpretation.