How to bring generative AI to oncology practice.

Area of Science:

• Oncology
• Artificial Intelligence
• Medical Informatics

Background:

• Generative artificial intelligence (AI), particularly large language models (LLMs), is increasingly relevant in oncology due to the field's reliance on text and data.
• Current adoption involves stand-alone chat models, with a progression towards more sophisticated systems.

Purpose of the Study:

• To review the current adoption of generative AI in oncology.
• To outline a practical pathway for integrating AI tools, from basic chat models to advanced agentic assistants.
• To identify concrete applications and constraints of AI in clinical oncology.

Main Methods:

• Review of current generative AI adoption in oncology.
• Conceptualization of a development trajectory for AI tools in healthcare.
• Identification of potential clinical use cases and technical/ethical challenges.

Main Results:

• Generative AI offers diverse applications including molecular tumor board synthesis, guideline-based grading, drafting radiology and pathology reports, and computable clinical trial matching.
• Key constraints include fragmented IT infrastructure, privacy concerns, data provenance, domain variability, and AI hallucinations.
• A phased adoption strategy is proposed, moving from drafting assistants to embedded tools and finally to event-driven agents.

Conclusions:

• Generative AI, especially LLMs, holds significant potential to augment oncology care by improving efficiency and supporting clinical decision-making.
• Successful integration requires addressing technical and ethical challenges and a carefully planned adoption trajectory.
• The ultimate goal is AI as a supportive learning assistant, enhancing routine care without replacing clinician judgment.