Automating Lung-RADS Categorization And Follow-Up Recommendations Using In-Context Learning With Large Language...

Area of Science:

• Radiology
• Artificial Intelligence
• Oncology

Background:

• Lung cancer is a leading cause of mortality, necessitating early detection through screening.
• Low-dose computed tomography (LDCT) screening is effective for high-risk individuals.
• Interpreting radiology reports for lung nodules is time-consuming and can be ambiguous, even with Lung-RADS.

Purpose of the Study:

• To develop an in-context learning framework using large language models (LLMs) to streamline lung nodule assessment.
• To identify an optimal LLM approach for accurate lung nodule categorization and management decisions.
• To provide robust and interpretable decision support for radiologists in lung cancer screening.

Main Methods:

• Utilized a large language model (LLM) framework with in-context learning.
• Focused on processing original radiology reports for lung nodule analysis.
• Aimed to generate Lung-RADS assessments and support clinical decision-making.

Main Results:

• The LLM framework demonstrated potential in categorizing lung nodules from reports.
• The study explored methods to enhance accuracy and interpretability of LLM outputs.
• The research focused on reducing radiologist workload in lung cancer screening.

Conclusions:

• LLMs offer a promising solution to challenges in interpreting lung cancer screening reports.
• This approach can enhance efficiency and accuracy in radiologist workflows.
• The framework supports timely and precise interventions for lung cancer patients.