Improving the Quality of Unstructured Cancer Data Using Large Language Models: A German Oncological Case Study

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Cancer is a leading global cause of death, necessitating robust epidemiological and clinical registration.
• Manual tumor documentation faces challenges due to heterogeneous medical data formats.
• Accurate cancer data is crucial for improving patient care, treatment efficacy, and scientific research.

Purpose of the Study:

• To explore the utility of Large Language Models (LLMs) in transforming unstructured medical reports into structured data.
• To assess the potential of LLMs for meeting the requirements of the German Basic Oncology Dataset.
• To evaluate the impact of LLM integration on cancer data quality and completeness.

Main Methods:

• Utilized Large Language Models (LLMs) to process and structure unstructured clinical cancer reports.
• Focused on converting free-text medical narratives into the standardized format of the German Basic Oncology Dataset.
• Explored integration strategies for LLMs within existing hospital data management systems and cancer registries.

Main Results:

• LLMs demonstrated significant potential in converting unstructured medical text to structured oncology data.
• Integration of LLMs can substantially improve the quality and completeness of cancer data collection.
• The findings support the use of LLMs for enhancing oncological data management.

Conclusions:

• LLMs offer a transformative solution for the challenges in manual cancer data documentation.
• Implementing LLMs can lead to more comprehensive and accurate cancer registries.
• This technology has the potential to revolutionize medical data processing in oncology and beyond.