AI Matches Treatment Choices in Precision Oncology Tests

Area of Science:

• Oncology
• Artificial Intelligence
• Genomics

Background:

• Precision oncology relies on integrating complex genomic data for treatment selection.
• Standard large language models (LLMs) face challenges in applying specialized medical knowledge accurately.
• There is a need for reliable AI tools to assist oncologists in data-driven treatment decisions.

Purpose of the Study:

• To evaluate the performance of an AI system in matching oncologists' treatment choices for genomics-guided therapy selection.
• To compare the AI system's accuracy against a standard large language model in a real-world clinical setting.

Main Methods:

• Development of an AI system utilizing structured oncology data.
• Testing the AI system on real clinical queries requiring genomics-guided therapy selection.
• Comparison of the AI system's treatment recommendations with those of oncologists and a standard LLM.

Main Results:

• The AI system successfully matched oncologists' treatment choices in over 90% of clinical queries.
• The AI system outperformed a standard large language model in this specific task.
• The AI system demonstrated high reliability in genomics-guided therapy selection.

Conclusions:

• AI systems powered by structured oncology data can reliably assist in precision oncology.
• Specialized AI platforms focused on specific tasks like genomics-guided therapy selection offer a practical deployment pathway.
• This approach validates the potential of AI to enhance clinical decision-making in oncology.